I'm a member of
Quicksearch
Codenews Search
Disclaimer
The individual owning this blog works for Oracle in Germany. The opinions expressed here are his own, are not necessarily reviewed in advance by anyone but the individual author, and neither Oracle nor any other party necessarily agrees with them.
Navigation
Categories
About this blog- Apple
- iPhone
- Aviation
- Bahn
- Begriffe des Lebens
- Blogosphere
- Braindump
- Business Travel
- CANBOX
- CeBIT
- CEC
- Computing
- del.icio.us
- Fundsache
- General
- HHOSUG
- i hate sundays
- Kochen
- Mindfuck
- Movies
- Music
- On Benchmarketing
- Persoenliches Pech
- Photographie
- policy of ...
- Privacy
- Recommended Blogs
- Security
- Solaris
- Strange Days
- Sun/Oracle
- suncec2007
- suncec2008
- Technology
- The IT Business
- Toolbox
- Travel
- Work
The Sequence
Saturday, April 10. 2010
Trackbacks
Trackback specific URI for this entry
No Trackbacks
Comments
Display comments as
(Linear | Threaded)
I feel like I am at a disadvantage here, when you obviously have a tendency to use your blog as a weapon to invoke an argument against someone who disagrees with your opinions because of your correlation with Oracle.
Having said that, business need does lead to application. But application does not lead to operating system(Oracle DB runs on Linux, Solaris, and a whole bunch of other OSes. So application is decoupled from OS choice) and operating system does not lead to hardware choices.(Linux and Solaris both run on x86, Power, Sparc...blah blah) So your chain of logic ends at "Business need leads to application".
Lastly, "Just a an example: Two systems were relatively easy to sell for me to new customers: The X45(0,4)0, because it was the only tier-1 system with that number of harddisks, and the X46(0,4)0 as it was the the only Tier-1 8-socket system for quite a time. So i was able to break into this already preagreed global contracts. I can't count the number of occurances, where a customer said to me "cool system, but it isn't in our book of standards" or "cool system, but we have already an agreement with vendor X/Y/Z for this class of system"."
It is amazing that you are still pumping X4600/X4640s considering that 4S Opteron 6100 offers 48 cores and 48 DIMMs and in 2U space in a Dell R815, or that Nehalem-EX is packing 64 threads in a 1U supermicro. And the Thumper and Thor isn't exactly good systems either, because you can always attach as many external JBODs to get the number of spindles you need. Let's be brutally honest here, you are pumping those obsolete gear because the profit margin is there, and your customers don't know any better.
Please, don't use your blog to attack personal opinions of your readers when your opinion is "paid for".
Having said that, business need does lead to application. But application does not lead to operating system(Oracle DB runs on Linux, Solaris, and a whole bunch of other OSes. So application is decoupled from OS choice) and operating system does not lead to hardware choices.(Linux and Solaris both run on x86, Power, Sparc...blah blah) So your chain of logic ends at "Business need leads to application".
Lastly, "Just a an example: Two systems were relatively easy to sell for me to new customers: The X45(0,4)0, because it was the only tier-1 system with that number of harddisks, and the X46(0,4)0 as it was the the only Tier-1 8-socket system for quite a time. So i was able to break into this already preagreed global contracts. I can't count the number of occurances, where a customer said to me "cool system, but it isn't in our book of standards" or "cool system, but we have already an agreement with vendor X/Y/Z for this class of system"."
It is amazing that you are still pumping X4600/X4640s considering that 4S Opteron 6100 offers 48 cores and 48 DIMMs and in 2U space in a Dell R815, or that Nehalem-EX is packing 64 threads in a 1U supermicro. And the Thumper and Thor isn't exactly good systems either, because you can always attach as many external JBODs to get the number of spindles you need. Let's be brutally honest here, you are pumping those obsolete gear because the profit margin is there, and your customers don't know any better.
Please, don't use your blog to attack personal opinions of your readers when your opinion is "paid for".
Well ... i can just ask you, to create your own blog, create own content and generate a large readership, when you feel at a disavantage.
However: I'm not paid to write this blog. Sorry ... there are people thinking that. The domain is private funded and not reimbursed, the server is privately funded and not remimbursed and i don't know how many man years of free-time i've invested into this blog. When i'm blog at day, it's often reimbursed to the company by working at night for company. With each new manager i have to fight again for the time. Sorry ... it seems easy to dismiss my opinion as a payed one, but the reality is different. Before being forced to write something i would rather just fall in a mode just writing about business travel or stopping this blog all together. And people that personally met me, know that ...
That said, i'm observing still a tendency in your comment to rapidly fall into technology when it's about business. That's the basic difference between our opinions, as you thing in technology terms, i have learned the hard way, that ultimately doesn't matter. It's important to know everything about the hardware, but it's even more important to understand the business side.
Even here you try to pitch some dell server. By using the example of the X4540 or the X4640 i wanted to describe the problem, that hardware decision isn't half as free as most people think. When a company has standardized on HP x86 servers, the Dell server can be the greatest server on the planet, they won't purchase it, because the costs of establishing new support processes, to establish new procuring processes outweight the cost advantages. Even when you convince that IT department, you will be stopped by the procurement department.
However there is a hole in this wall: Servers that are different and where customers see an business justification to use something different. While i'm thinking you are somewhat right about the point that the 8 socket market will see problems from 4 socket systems, the x4540 had/has still a very valid advantage justifying it's purchase even in pure HP/IBM/Dell account. But this article isn't about pitching Sun equipment so i won't talk here about Sun.
Oracle DB isn't an application. It's middleware. A SQL database alone solves no business problem. It enables an application to do something. To be more precise: "Business need leads to application, application leads to middleware, middleware and application leads to os, os leads to hardware architecture, hardware architecture to hardware components".
Just a comment to your point of: You can choose what ever you want for your application. There is a vast amount of dependencies and constraints. Even when there is Linux on the Box, it's almost everytime SuSE or RedHat on x86. Not CentOS or Debian. AIX leads automatically to Power, whereas the appliaction support matrices of supported operating systems leads to AIX with Power, as application vendor support for Linux on Power is practically unknown. Even in the large portfolios of the large ERP software vendors, there is software just running on Windows for Example, but not on unix, and software that is just running on unix but not on windows.
IT and business is a world full of constraints, full of political decisions ... hardware is nice, but only a small part. This is what i'm trying to explain to you since several comments for various articles.
However: I'm not paid to write this blog. Sorry ... there are people thinking that. The domain is private funded and not reimbursed, the server is privately funded and not remimbursed and i don't know how many man years of free-time i've invested into this blog. When i'm blog at day, it's often reimbursed to the company by working at night for company. With each new manager i have to fight again for the time. Sorry ... it seems easy to dismiss my opinion as a payed one, but the reality is different. Before being forced to write something i would rather just fall in a mode just writing about business travel or stopping this blog all together. And people that personally met me, know that ...
That said, i'm observing still a tendency in your comment to rapidly fall into technology when it's about business. That's the basic difference between our opinions, as you thing in technology terms, i have learned the hard way, that ultimately doesn't matter. It's important to know everything about the hardware, but it's even more important to understand the business side.
Even here you try to pitch some dell server. By using the example of the X4540 or the X4640 i wanted to describe the problem, that hardware decision isn't half as free as most people think. When a company has standardized on HP x86 servers, the Dell server can be the greatest server on the planet, they won't purchase it, because the costs of establishing new support processes, to establish new procuring processes outweight the cost advantages. Even when you convince that IT department, you will be stopped by the procurement department.
However there is a hole in this wall: Servers that are different and where customers see an business justification to use something different. While i'm thinking you are somewhat right about the point that the 8 socket market will see problems from 4 socket systems, the x4540 had/has still a very valid advantage justifying it's purchase even in pure HP/IBM/Dell account. But this article isn't about pitching Sun equipment so i won't talk here about Sun.
Oracle DB isn't an application. It's middleware. A SQL database alone solves no business problem. It enables an application to do something. To be more precise: "Business need leads to application, application leads to middleware, middleware and application leads to os, os leads to hardware architecture, hardware architecture to hardware components".
Just a comment to your point of: You can choose what ever you want for your application. There is a vast amount of dependencies and constraints. Even when there is Linux on the Box, it's almost everytime SuSE or RedHat on x86. Not CentOS or Debian. AIX leads automatically to Power, whereas the appliaction support matrices of supported operating systems leads to AIX with Power, as application vendor support for Linux on Power is practically unknown. Even in the large portfolios of the large ERP software vendors, there is software just running on Windows for Example, but not on unix, and software that is just running on unix but not on windows.
IT and business is a world full of constraints, full of political decisions ... hardware is nice, but only a small part. This is what i'm trying to explain to you since several comments for various articles.
The issue I have with you is that you use your blog posts to specifically target individuals: SteveA, me, Tim "BlueToTheBones"....that is insulting, especially you are the one blinded by your own correlation with Oracle. No, you are not paid to write the blog, but nevertheless, as UX-Admin pointed out, Oracle is the one putting bread on your table right now, and most of the hardware choices you pumped on this blog have been proven to be suboptimal choices. ie, X4640, F5100, Rainbow falls, Sun 7410s, Stec, where there are far superior choices on the open market for fraction of cost and often with higher performances.
I would like to conclude this argument by saying this:
"where a customer said to me "cool system, but it isn't in our book of standards" or "cool system, but we have already an agreement with vendor X/Y/Z for this class of system"
When a customer tells you that, the German translation is this: "Sun gear sucks balls for the money, all we need is Solaris, and that's only because it is free. Let's face it, the only server from the entire Sun portfolio that possibly pass the litmus test is the X4170 and even then, you are paying a significant premium over DL360G6 to use it.
I hope within a few months, I can hear from you about OpenSolaris 2010.1H.
I would like to conclude this argument by saying this:
"where a customer said to me "cool system, but it isn't in our book of standards" or "cool system, but we have already an agreement with vendor X/Y/Z for this class of system"
When a customer tells you that, the German translation is this: "Sun gear sucks balls for the money, all we need is Solaris, and that's only because it is free. Let's face it, the only server from the entire Sun portfolio that possibly pass the litmus test is the X4170 and even then, you are paying a significant premium over DL360G6 to use it.
I hope within a few months, I can hear from you about OpenSolaris 2010.1H.
Well ... Steve A. got his special mention for faking a discussion, Bluetothebone ... well ... it's pretty much known in the industry that TPM is extremely positive in regard of Big Blue, in regard of IBM, despite calling himself a journalist.
That said, this comment shows, why a discussion is futile with you. We discussed many of stuff already, but you still seem to be convinced that speeds and feeds are the only thing in the industry. We already had that discussion, again and again. The STEC discussion, the 7000 discussion, i can't count them but i'm getting more and more the impression that i'm talking with a wall. I'm somewhat tired of this ...
I think, it could be an insightful experience for you to work for a hardware vendor for a while, regardless if it's a storage or a server vendor, i think it would be even insightful for you to work within the IT department of a non-internet related company: That would help you to see the world looks differently when you look from other perspectives, that hardware is just one, most of the times a secondary part of IT.
By the way: In the series of the nehalem based systems, it's the X4275 that is most interesting, not the X4170 ... but's that just my opinion.
That said, this comment shows, why a discussion is futile with you. We discussed many of stuff already, but you still seem to be convinced that speeds and feeds are the only thing in the industry. We already had that discussion, again and again. The STEC discussion, the 7000 discussion, i can't count them but i'm getting more and more the impression that i'm talking with a wall. I'm somewhat tired of this ...
I think, it could be an insightful experience for you to work for a hardware vendor for a while, regardless if it's a storage or a server vendor, i think it would be even insightful for you to work within the IT department of a non-internet related company: That would help you to see the world looks differently when you look from other perspectives, that hardware is just one, most of the times a secondary part of IT.
By the way: In the series of the nehalem based systems, it's the X4275 that is most interesting, not the X4170 ... but's that just my opinion.
I am not even in the IT industry, why do you think I should work in the IT industry? My interest in high performance computing and storage is a just a small building block for what I really do.
You are tired of our discussions because you are running out of points. When you lose an argument, you tend to cop out by saying "it is futile".
And you keep on saying hardware is not important. This is clearly false. Sun failed because SPARC failed to be competitive against x86-64, dragging Solaris down with it.
I tend not to be vocal unless you are pumping shit, or insulting me or other people who didn't deserve it. So you know the way to shut me up:
1. Don't pump shit.
2. Don't be Oracle's nonsense PR machine. The best way to stop the "OpenSolaris is dead" FUD is to release the damn distro.
You are tired of our discussions because you are running out of points. When you lose an argument, you tend to cop out by saying "it is futile".
And you keep on saying hardware is not important. This is clearly false. Sun failed because SPARC failed to be competitive against x86-64, dragging Solaris down with it.
I tend not to be vocal unless you are pumping shit, or insulting me or other people who didn't deserve it. So you know the way to shut me up:
1. Don't pump shit.
2. Don't be Oracle's nonsense PR machine. The best way to stop the "OpenSolaris is dead" FUD is to release the damn distro.
I have a budget of 24 hours per day. Don't you think that i have better to do than discussing with someone, who just told, that he doesn't even work in IT and having a tight focus on HPC with completely different market dynamics than commercial IT? I've already suspected from all your comments in this blog, that your perspective to IT is somewhat extremely focussed ....
I could discuss with you forever, but you don't show even the slighest sign that you are even thinking about the stuff i'm telling to you. So it's wasted time and so i'm tired of this.
Why do i think, you should work in that business, at least for a while? That's easy - because you then might think different about a lot of development in IT and may get to the conclusion, that the world isn't that simple as you paint it here again and again without showing any will even to slightly admit, that there may other objectives in other areas of IT that may have as same validity as your point of view.
I could discuss with you forever, but you don't show even the slighest sign that you are even thinking about the stuff i'm telling to you. So it's wasted time and so i'm tired of this.
Why do i think, you should work in that business, at least for a while? That's easy - because you then might think different about a lot of development in IT and may get to the conclusion, that the world isn't that simple as you paint it here again and again without showing any will even to slightly admit, that there may other objectives in other areas of IT that may have as same validity as your point of view.
Let me tell you a story:
A business man goes to a shoe shiner to get his shoes shined. The shoe shiner uses Sun branded shoe polish and the business man isn't very happy about that and asks the shoe shiner: "Why don't you use Dell or HP Shoe polish?"
The shoe shiner says, "why don't you become a shoe shiner yourself for a year to tell me which brand of shoe polish I can or cannot use? Plus this isn't about what brand of shoe polish I use, it is about my service of shinning your shoes"
Joerg, you are a good guy, unfortunately, as long as you are at Oracle, your field of vision would be understandably restricted. I feel your pain, it is tough defending a premium product that doesn't have premium performance.
Let the conversation just stop here. I got other things to do.
A business man goes to a shoe shiner to get his shoes shined. The shoe shiner uses Sun branded shoe polish and the business man isn't very happy about that and asks the shoe shiner: "Why don't you use Dell or HP Shoe polish?"
The shoe shiner says, "why don't you become a shoe shiner yourself for a year to tell me which brand of shoe polish I can or cannot use? Plus this isn't about what brand of shoe polish I use, it is about my service of shinning your shoes"
Joerg, you are a good guy, unfortunately, as long as you are at Oracle, your field of vision would be understandably restricted. I feel your pain, it is tough defending a premium product that doesn't have premium performance.
Let the conversation just stop here. I got other things to do.
>suboptimal choices. ie, bla bla , Sun 7410s, Stec,
Do you run 7410's with STEC Drives? No?
I do, and I am very happy with its performance, while being substantially cheaper (and still faster) than what other vendors have.
Do you run 7410's with STEC Drives? No?
I do, and I am very happy with its performance, while being substantially cheaper (and still faster) than what other vendors have.
You are right I don't run Stecs and 7410s. You are happy about the performance because you aren't intelligent enough to know the alternatives. Philosophers often say that happiness is the differential between reality and what you expect. In your case, you expectation is too low because you are not aware what Sun really sold you.
BTW, the J4400 JBOD is a rebadge of DataOn DNS-1400DM
http://www.dataonstorage.com/dataon-products/dns-1400-4u-sas-to-sas-sata-jbod-storage.html
The 7410 is just a X4440 with a really slow SAS 1.0 HBA.
You could have cut your billl by 70% if you just ordered the X4440 server with the same specs and the J4400 or DataOn DNS-1400DM and put NexentaStor 3.0 on your system and get the same performance.
As to Stec JesusIOPs, significant number of SSDs outshine its $300/GB price tag, X25-E, Micron C300, SandForce, Vertex2-EX. As to Stec Mach8 L2ARC, all those above SSDs will beat it in performance at 1/3 of the price.
But I can't tell you how to waste your money. So keep on being happy in your Sun washed mind.
BTW, the J4400 JBOD is a rebadge of DataOn DNS-1400DM
http://www.dataonstorage.com/dataon-products/dns-1400-4u-sas-to-sas-sata-jbod-storage.html
The 7410 is just a X4440 with a really slow SAS 1.0 HBA.
You could have cut your billl by 70% if you just ordered the X4440 server with the same specs and the J4400 or DataOn DNS-1400DM and put NexentaStor 3.0 on your system and get the same performance.
As to Stec JesusIOPs, significant number of SSDs outshine its $300/GB price tag, X25-E, Micron C300, SandForce, Vertex2-EX. As to Stec Mach8 L2ARC, all those above SSDs will beat it in performance at 1/3 of the price.
But I can't tell you how to waste your money. So keep on being happy in your Sun washed mind.
facepalm Hey, you can't whine about being insulted at the beginning of the thread, and then insult a reader.
What exactly was the insult?
Mika stated that he's happy with his 7410. My conjecture was that he's only happy because he's not well informed of alternatives and that his expectation is too low.
Then I provided a list of gears that is exactly the same as 7410, only to be labeled "Radio Shack" gear.
I know an insult when I see one, trust me, you should read beyond the lines.
Mika stated that he's happy with his 7410. My conjecture was that he's only happy because he's not well informed of alternatives and that his expectation is too low.
Then I provided a list of gears that is exactly the same as 7410, only to be labeled "Radio Shack" gear.
I know an insult when I see one, trust me, you should read beyond the lines.
Allow me to cite from your comment:
"You are happy about the performance because you aren't intelligent enough to know the alternatives"
You didn't wrote "not informed" ... you wrote "not intelligent" .... that easily qualifies as insult in every culture. That's your basic problem: You consider every decision that doesn't fit in your perspective of the world as "non-intelligent", "dumb" or in the case of hardware vendors as "greedy".
I've tried to explain to you that the world isn't that easy ... but somehow you don't see that other people may get to different conclusions. And this behaviour, TS, easily qualifies as arrogance.
"You are happy about the performance because you aren't intelligent enough to know the alternatives"
You didn't wrote "not informed" ... you wrote "not intelligent" .... that easily qualifies as insult in every culture. That's your basic problem: You consider every decision that doesn't fit in your perspective of the world as "non-intelligent", "dumb" or in the case of hardware vendors as "greedy".
I've tried to explain to you that the world isn't that easy ... but somehow you don't see that other people may get to different conclusions. And this behaviour, TS, easily qualifies as arrogance.
Yes, I am arrogant.
Look at recent Sun hardware releases: F5100, Sun 7410, F20 SSD. If there is one trend, it is this: Sun is taking every opportunity to chop your arm and legs off. That's why Sun failed in the first place.
Think of me as a vigilante who informs the public that it is fucking stupid to buy a 1TB spindle from Sun for $800 when the open market price is $200 for Hitachi UltraStars, and $150 for Seagate Barracuda ES.2s.
BTW, I want to take one last stab at STEC:
http://www.stec-inc.com/product/mach8.php
100MB/sec read/write, 5000 IOPs read, 100! IOPs writes. Shit, pretty much every SSD on the market except JMicron based ones can beat those specs, yet Sun charges $5K for a 100GB Mach8 or $50/GB, when Intel X25-E can do 250M read/170MB writes, 33K read IOPs, 3.3K write IOPs at $12/GB. And I don't want to talk about JesusIOPs at $300+/GB when it is just a dumb ass supercapacitor.
You have to be licking someone's balls to be this retarded when it comes to hardware choices.
Look at recent Sun hardware releases: F5100, Sun 7410, F20 SSD. If there is one trend, it is this: Sun is taking every opportunity to chop your arm and legs off. That's why Sun failed in the first place.
Think of me as a vigilante who informs the public that it is fucking stupid to buy a 1TB spindle from Sun for $800 when the open market price is $200 for Hitachi UltraStars, and $150 for Seagate Barracuda ES.2s.
BTW, I want to take one last stab at STEC:
http://www.stec-inc.com/product/mach8.php
100MB/sec read/write, 5000 IOPs read, 100! IOPs writes. Shit, pretty much every SSD on the market except JMicron based ones can beat those specs, yet Sun charges $5K for a 100GB Mach8 or $50/GB, when Intel X25-E can do 250M read/170MB writes, 33K read IOPs, 3.3K write IOPs at $12/GB. And I don't want to talk about JesusIOPs at $300+/GB when it is just a dumb ass supercapacitor.
You have to be licking someone's balls to be this retarded when it comes to hardware choices.
As much as I disagree with Joerg on almost any subject I must give him credit for his position on storage.
If you, TS, wonder why an enterprise storage system costs more than the individual parts you should take a look at an interoperability matrix if you ever get your hands on one. That matrix documents tested configurations for a storage system together with switch type and firmware, host OS and patch level, host HBA, HBA-driver, HBA firmware and multipath drivers. The document for EMC Symmetrix is 7213 pages long. These tested configurations comes in addition to the plethora of tests performed on the arry itself. I guess that is where SUN wants to be with their own range of storage. That is supply systems that don't crash and burn or craps out after a firmware upgrade like I have seen so many times.
This is why most serious storage companies are behind regarding the latest technology advances in chip design, disks and memory. This is why feature X takes a long time to implement.
And like storage companies, storage administrators are conservative. They have respect for the things they do and are careful. They don't act on wild ideas on how to improve performance.
That is why I don't think you are responsible for anything remotely critical when it comes to storage. In fact, your attitude, your overly self confident behavior and your Tomshardware approach to this subject would make you a liability in any organization that cherish its data.
If you, TS, wonder why an enterprise storage system costs more than the individual parts you should take a look at an interoperability matrix if you ever get your hands on one. That matrix documents tested configurations for a storage system together with switch type and firmware, host OS and patch level, host HBA, HBA-driver, HBA firmware and multipath drivers. The document for EMC Symmetrix is 7213 pages long. These tested configurations comes in addition to the plethora of tests performed on the arry itself. I guess that is where SUN wants to be with their own range of storage. That is supply systems that don't crash and burn or craps out after a firmware upgrade like I have seen so many times.
This is why most serious storage companies are behind regarding the latest technology advances in chip design, disks and memory. This is why feature X takes a long time to implement.
And like storage companies, storage administrators are conservative. They have respect for the things they do and are careful. They don't act on wild ideas on how to improve performance.
That is why I don't think you are responsible for anything remotely critical when it comes to storage. In fact, your attitude, your overly self confident behavior and your Tomshardware approach to this subject would make you a liability in any organization that cherish its data.
Do you not understand the concept of "copy exact"? Not that what you are copying is the most optimal solution, but you really don't have anything against a copy exact 7410.
http://www.google.com/search?hl=en&q=define%3Aintelligent&btnG=Search
It could mean possessing sound knowledge; "well-informed readers"
Intelligent people are well connected with information, but it doesn't make them smart because smart people need to process the "intelligence" in a right way in order to profit.
But truthfully, Mika doesn't really know jack squat about storage anyways.
It could mean possessing sound knowledge; "well-informed readers"
Intelligent people are well connected with information, but it doesn't make them smart because smart people need to process the "intelligence" in a right way in order to profit.
But truthfully, Mika doesn't really know jack squat about storage anyways.
0. Sorry, TS ... i've was really lenient to you and your behaviour, but the line will be drawn here. Act your age or you will the first person specifically mentioned in the matching spam filter set. I will non't tolerate your behavior any longer.
Regarding your insult: Don't flunk out ... just look into the merriam-webster about the meaning of the word intelligent ... http://www.merriam-webster.com/dictionary/intelligent
1. I thought you were just arrogant, but given the last few mails you look you have a severe savior complex. You seem to be very inclined that your wisdom will enlight the world.
2. Don't consider yourself as the single source of truth. Especially when you have really no first hand experience with gear. When you compare numbers of data sheets, be sure that they really measure the same, measured on the same way. Do you really think that "100 Transactions" is the same than "3300 IOPS"?
3. In the last few days you've destroyed every thought that your comments are driven by insight to the things you are talking about. Maybe my insight is narrowed by working at Sun/Oracle, but your understanding of IT is narrowed to a rather small part of IT when you compare the revenues in HPC and in commercial computing. Your comment that you don't work in enterprise IT and just interested in HPC is a sign of that.
4. Regarding your harddisk price comment: Harddisks costs almost the same at all the Tier-1 vendors (Dell,HP,Sun/Oracle,IBM), and there are many reasons why all those vendors calls for such a price. I've explained this, i won't repeat it. BTW: Did you know that Dell has started to enforce the usage of Dell disks with some of their RAID controllers?
5. Given the fact, that you never worked for a hardware vendor nor was in the position to responsibly decide about a hardware, which availability is pivotal to the ability of hundred or thousands of people to earn money and to deliver services, it's understandable that you have no insight into the mindset of people having to decide about such hardware.
6. You write, that firmware problems or software problems are non-existent. That's interesting. Every knowledgeable person in computer science knows that it's impossible to create a computer program or it's representation in hardware without errors. I assume that you know how many independent parts of software system runs in a server, how many logic chips with high complexity are running in such a chip. Do you really think, that even the chips of your beloved Intel Cooperation are excempt from this? Don't believe it? Look how many fixes and workarounds are in open-source operating systems just to fix the short falls of hardware. Just think of the problems of Linuxes and Open(Solaris) with some Broadcom chips with a firmware version 5.x whereas firmware versions 4.x work.
The only way to rule out such problems is testing, massive testing, throughoutly testing. In exactly the workload you want to use the system. And that's the reason why the qualification process is really important. Try to go your boss and explain him, that you've lost data because of a rare incompatibility in SAS protocols? Or problems in a certain set protocol conversion or tunneling protocols? When important data is at risk you have even to check if your PCIe bus switch isn't corrupting data when extremely loaded.
Your mindset is about assuming that everything will works as defined. For HPC this may be the right approach, when you can recompute everything after getting the dataset from backup. Enterprise computing is about knowing that your gear works as designed with a high degree of assurance.
Would you guarantee (important: not assume) that your beloved enthusiasts SSD don't start to shred data when under extreme load? That the controller just won't stop to work when under high load. Do you know the exact behaviour of the enthusiasts SSD under extreme load in regard of their performance characteristics, do you guarantee that those disks doesn't run in some unforseen performance characteristics when used at extremely high load?
And Mikas mindset is about providing services to million of peoples. The costs of a STEC SSD are infinitely small compared to the costs when his company stands in the newspaper, because of a mayor fuckup in storing user data. The costs of paying Sun to take care of this problem are infinitely small compared to costs of a major outage for millions of users.
It's a different mindset, and i don't accuse you of not having this mindset, as you never worked with this kind of responsibility. But in this light: Don't pretend that you know everything better than everbody, and don't insult people having a different mindset created out of the situation, that they have such responsibilities.
Regarding your insult: Don't flunk out ... just look into the merriam-webster about the meaning of the word intelligent ... http://www.merriam-webster.com/dictionary/intelligent
1. I thought you were just arrogant, but given the last few mails you look you have a severe savior complex. You seem to be very inclined that your wisdom will enlight the world.
2. Don't consider yourself as the single source of truth. Especially when you have really no first hand experience with gear. When you compare numbers of data sheets, be sure that they really measure the same, measured on the same way. Do you really think that "100 Transactions" is the same than "3300 IOPS"?
3. In the last few days you've destroyed every thought that your comments are driven by insight to the things you are talking about. Maybe my insight is narrowed by working at Sun/Oracle, but your understanding of IT is narrowed to a rather small part of IT when you compare the revenues in HPC and in commercial computing. Your comment that you don't work in enterprise IT and just interested in HPC is a sign of that.
4. Regarding your harddisk price comment: Harddisks costs almost the same at all the Tier-1 vendors (Dell,HP,Sun/Oracle,IBM), and there are many reasons why all those vendors calls for such a price. I've explained this, i won't repeat it. BTW: Did you know that Dell has started to enforce the usage of Dell disks with some of their RAID controllers?
5. Given the fact, that you never worked for a hardware vendor nor was in the position to responsibly decide about a hardware, which availability is pivotal to the ability of hundred or thousands of people to earn money and to deliver services, it's understandable that you have no insight into the mindset of people having to decide about such hardware.
6. You write, that firmware problems or software problems are non-existent. That's interesting. Every knowledgeable person in computer science knows that it's impossible to create a computer program or it's representation in hardware without errors. I assume that you know how many independent parts of software system runs in a server, how many logic chips with high complexity are running in such a chip. Do you really think, that even the chips of your beloved Intel Cooperation are excempt from this? Don't believe it? Look how many fixes and workarounds are in open-source operating systems just to fix the short falls of hardware. Just think of the problems of Linuxes and Open(Solaris) with some Broadcom chips with a firmware version 5.x whereas firmware versions 4.x work.
The only way to rule out such problems is testing, massive testing, throughoutly testing. In exactly the workload you want to use the system. And that's the reason why the qualification process is really important. Try to go your boss and explain him, that you've lost data because of a rare incompatibility in SAS protocols? Or problems in a certain set protocol conversion or tunneling protocols? When important data is at risk you have even to check if your PCIe bus switch isn't corrupting data when extremely loaded.
Your mindset is about assuming that everything will works as defined. For HPC this may be the right approach, when you can recompute everything after getting the dataset from backup. Enterprise computing is about knowing that your gear works as designed with a high degree of assurance.
Would you guarantee (important: not assume) that your beloved enthusiasts SSD don't start to shred data when under extreme load? That the controller just won't stop to work when under high load. Do you know the exact behaviour of the enthusiasts SSD under extreme load in regard of their performance characteristics, do you guarantee that those disks doesn't run in some unforseen performance characteristics when used at extremely high load?
And Mikas mindset is about providing services to million of peoples. The costs of a STEC SSD are infinitely small compared to the costs when his company stands in the newspaper, because of a mayor fuckup in storing user data. The costs of paying Sun to take care of this problem are infinitely small compared to costs of a major outage for millions of users.
It's a different mindset, and i don't accuse you of not having this mindset, as you never worked with this kind of responsibility. But in this light: Don't pretend that you know everything better than everbody, and don't insult people having a different mindset created out of the situation, that they have such responsibilities.
Wow, Joerg: long response!
0. You should consider fortunate that I am participating on your blog. My age is irrelevant when it comes to logical discussions. Obviously you can't handle it when someone rips your company's product apart. As I have said before, that as long as you are at Oracle, you should consider your own opinions about Oracle gears biased.
1. I don't pretend that my information is saving the world, just reducing the number of pigs falling into your lies.
2. Transactions is probably not the same as IOPs, but none the less, without definition of "transaction" from STEC, the fact remains that 100MB/read 100MB/write and 100 write transactions is inferior than most of SSDs on the market. Prove me wrong.
3. No
4. Yes, I am aware Dell started requiring Dell branded HDDs in their H800 series controllers. That's why people are still using SAS5E and SAS6E and HP P411s. Dell will lose sales because of that. I am aware of how much Netapp and EMC charge for a spindle. The point is Sun's Open Storage was supposed to be different. Mike Shapiro mentioned about this in a talk, unfortunately, Sun is just another Tier1 vendor, only waving the "Open Source" storage flag. The true implementation is the bastard child. Wait and see.
5. Read the story about the shoe shiner again. That story is pretty insulting, but you don't get it.
6. I should have picked my words more carefully. What I meant about the "nonexistent" firmware problem is that the hardware HBA is the same, and you should upgrade to the latest firmware, and if there is still problem in the firmware, your official 7410 would have had the same problem since it is EXACTLY the same gear.
"Would you guarantee (important: not assume) that your beloved enthusiasts SSD don't start to shred data when under extreme load? That the controller just won't stop to work when under high load. Do you know the exact behaviour of the enthusiasts SSD under extreme load in regard of their performance characteristics, do you guarantee that those disks doesn't run in some unforseen performance characteristics when used at extremely high load? "
The proper question is does it matter? You have given a ton of negatives against the X25-e. My question to you is this: given that L2ARC is not persistent, does it matter if X25-e corrupts a few bits? It is a small cache miss and ZFS reads from the disks anyways. You also keep on hitting that X25-e's lack of supercapacitor. My question again for you: do it matter as L2ARC? It is not persistent. A reboot would have whipped it clean. Plus even if L2ARC somehow managed to become persistent, missing a few last transactions would just mean a little cache miss. Nothing major. You can use the X25-E as ZILs too if the HBA you used has a battery or flash on it.
So the best response you have for the STEC Mach8 is that "transaction" doesn't mean IOPs. So in order to continue, you should come back with a definition of transaction from STEC's glossary because the way I look at it 100MB read is still pretty weak for a L2ARC device.
Lastly, don't assume you know about my mindset. If you had a freaking clue why I am on this blog, you wouldn't have defended your hopeless position with that much effort.
0. You should consider fortunate that I am participating on your blog. My age is irrelevant when it comes to logical discussions. Obviously you can't handle it when someone rips your company's product apart. As I have said before, that as long as you are at Oracle, you should consider your own opinions about Oracle gears biased.
1. I don't pretend that my information is saving the world, just reducing the number of pigs falling into your lies.
2. Transactions is probably not the same as IOPs, but none the less, without definition of "transaction" from STEC, the fact remains that 100MB/read 100MB/write and 100 write transactions is inferior than most of SSDs on the market. Prove me wrong.
3. No
4. Yes, I am aware Dell started requiring Dell branded HDDs in their H800 series controllers. That's why people are still using SAS5E and SAS6E and HP P411s. Dell will lose sales because of that. I am aware of how much Netapp and EMC charge for a spindle. The point is Sun's Open Storage was supposed to be different. Mike Shapiro mentioned about this in a talk, unfortunately, Sun is just another Tier1 vendor, only waving the "Open Source" storage flag. The true implementation is the bastard child. Wait and see.
5. Read the story about the shoe shiner again. That story is pretty insulting, but you don't get it.
6. I should have picked my words more carefully. What I meant about the "nonexistent" firmware problem is that the hardware HBA is the same, and you should upgrade to the latest firmware, and if there is still problem in the firmware, your official 7410 would have had the same problem since it is EXACTLY the same gear.
"Would you guarantee (important: not assume) that your beloved enthusiasts SSD don't start to shred data when under extreme load? That the controller just won't stop to work when under high load. Do you know the exact behaviour of the enthusiasts SSD under extreme load in regard of their performance characteristics, do you guarantee that those disks doesn't run in some unforseen performance characteristics when used at extremely high load? "
The proper question is does it matter? You have given a ton of negatives against the X25-e. My question to you is this: given that L2ARC is not persistent, does it matter if X25-e corrupts a few bits? It is a small cache miss and ZFS reads from the disks anyways. You also keep on hitting that X25-e's lack of supercapacitor. My question again for you: do it matter as L2ARC? It is not persistent. A reboot would have whipped it clean. Plus even if L2ARC somehow managed to become persistent, missing a few last transactions would just mean a little cache miss. Nothing major. You can use the X25-E as ZILs too if the HBA you used has a battery or flash on it.
So the best response you have for the STEC Mach8 is that "transaction" doesn't mean IOPs. So in order to continue, you should come back with a definition of transaction from STEC's glossary because the way I look at it 100MB read is still pretty weak for a L2ARC device.
Lastly, don't assume you know about my mindset. If you had a freaking clue why I am on this blog, you wouldn't have defended your hopeless position with that much effort.
So enlighten me: Why do i have the great luck, that you shine with your great wisdom on my small, unworthy blog? 
People either get it or they don't. If you have to ask, you don't get it.
Because you try to save the computing world from me ? Don't let me die unenlighted
Don't let me die unenlighted
Why do you think your enlightenment upon death depends on another human being?
I'm just curious, why you thing that you are here with a mission, why it's so important that you are reading in my blog Savior complex ? Perhaps you should really get professional help
Savior complex ? Perhaps you should really get professional help
I am here because the open storage movement is significant and will have ramifications beyond your imaginations. As I said before, this movement is about family rivalry, the war between the official descendant vs the bastard child.
Don't confuse the psychology of a vigilante and the psychology of savior complex. A person who suffers the savior complex helps the man in trouble, but a vigilante strikes back at the villain who is causing the trouble.
Don't confuse the psychology of a vigilante and the psychology of savior complex. A person who suffers the savior complex helps the man in trouble, but a vigilante strikes back at the villain who is causing the trouble.
facepalm
Have you considered the possibility that an AssPalm is more creative for you?
'TS' stands for 'The Savior', natch.
TS I think that it is enough.
I don't agree with many comments of Joerg.
but your attitude isn't any good.
It is the blog of Joerg, he has done a hard work, and I think that you have to respect him and its blog.
I think that most people is smart and can decide one thing or other reading the comments posted on blog.
I don't agree with many comments of Joerg.
but your attitude isn't any good.
It is the blog of Joerg, he has done a hard work, and I think that you have to respect him and its blog.
I think that most people is smart and can decide one thing or other reading the comments posted on blog.
You are right Jose. I think it is enough information exchange for a day.
The point was that Joerg used his blog to specifically tell the world why my comments were wrong. Why do you not expect a rigorous response from me? It is not the first time this happened. I think this is the 3rd time Joerg publicly involved me into a lengthy argument to pump/defend Oracle hardware. Let's not forget the numerous BlueToTheBones and SteveA fiascos.
The point was that Joerg used his blog to specifically tell the world why my comments were wrong. Why do you not expect a rigorous response from me? It is not the first time this happened. I think this is the 3rd time Joerg publicly involved me into a lengthy argument to pump/defend Oracle hardware. Let's not forget the numerous BlueToTheBones and SteveA fiascos.
LOL! Next you will tell me to go to Radioshack.
Storage is not only about price and performance, it is also about availability and reliability (even more so).
While Nexenta does interesting things, they are not able to provide me with the same level of support and trust.
Obviously, you don't understand what the SSD Disks in the Sun Storage 7000 do, as you are talking about $ per GB. What I need is $ per IO. And regarding this, the STECs are doing a great job. I consider their price even as damn cheap.
Regarding your "insult". I know very well what Sun sold us, as this system has replaced a competitors system. We have a direct comparison, which you obviously don't have.
Overall the Sun Storage 7000 has the best mix between Price/GB, Price/IO, Reliabilty, Availabilty, Features, Support(costs) of all Unified Storages Systems out there.
And now go back playing with your Radioshack-Storage-system.
Storage is not only about price and performance, it is also about availability and reliability (even more so).
While Nexenta does interesting things, they are not able to provide me with the same level of support and trust.
Obviously, you don't understand what the SSD Disks in the Sun Storage 7000 do, as you are talking about $ per GB. What I need is $ per IO. And regarding this, the STECs are doing a great job. I consider their price even as damn cheap.
Regarding your "insult". I know very well what Sun sold us, as this system has replaced a competitors system. We have a direct comparison, which you obviously don't have.
Overall the Sun Storage 7000 has the best mix between Price/GB, Price/IO, Reliabilty, Availabilty, Features, Support(costs) of all Unified Storages Systems out there.
And now go back playing with your Radioshack-Storage-system.
Yes, the only thing you have against Nexenta is that "you don't think you trust them".
I don't know how much you paid for your 7410. What I am saying is this:
Go order a x4440+ DataOn DNS1400-DM and put the same drives and even the same SSDs in there, you would have the exact same hardware at 70% off. Care to calculate $/per IO?
"Overall the Sun Storage 7000 has the best mix between Price/GB, Price/IO, Reliabilty, Availabilty, Features, Support(costs) of all Unified Storages Systems out there." is a false marketing statement.
Be happy with your 7410.
I am happy with my radio-shack gear(dell/hp) and if Oracle didn't buy Sun out, you would have owned Dumpster-shack gear.
I don't know how much you paid for your 7410. What I am saying is this:
Go order a x4440+ DataOn DNS1400-DM and put the same drives and even the same SSDs in there, you would have the exact same hardware at 70% off. Care to calculate $/per IO?
"Overall the Sun Storage 7000 has the best mix between Price/GB, Price/IO, Reliabilty, Availabilty, Features, Support(costs) of all Unified Storages Systems out there." is a false marketing statement.
Be happy with your 7410.
I am happy with my radio-shack gear(dell/hp) and if Oracle didn't buy Sun out, you would have owned Dumpster-shack gear.
>Yes, the only thing you have against Nexenta is that "you don't think
>you trust them".
Maybe you have never heard about the term "validated configuration". Every serious storage vendor exactly specifies, which components, down to the firmware, are supported.
If a customer has a problem, the vendor knows exactly what's out there in the field.
Now, if we go back to your Radioshack system, what happens if something goes wrong? Let's say you have a problem with your high-speed SAS adapter (driver, firmware etc.). Can you call Nexenta for support, when you have a problem. Maybe they even haven't heard of the type of hba you own?
This is one of the main reasons, I don't trust them.
Maybe in other environments (like HPC), this model is ok. But here, where we have millions of customers accessing the system, this just doesn't work out.
>you trust them".
Maybe you have never heard about the term "validated configuration". Every serious storage vendor exactly specifies, which components, down to the firmware, are supported.
If a customer has a problem, the vendor knows exactly what's out there in the field.
Now, if we go back to your Radioshack system, what happens if something goes wrong? Let's say you have a problem with your high-speed SAS adapter (driver, firmware etc.). Can you call Nexenta for support, when you have a problem. Maybe they even haven't heard of the type of hba you own?
This is one of the main reasons, I don't trust them.
Maybe in other environments (like HPC), this model is ok. But here, where we have millions of customers accessing the system, this just doesn't work out.
Mika:
Do you know the difference between me and you? When I tell you how to source a 7410 system from DataON and Sun X4440, you call it "Radio-Shack" gear, but you don't even know that your so called "Super doper Sun" gear is a "Radio-Shack rebadge".
"Now, if we go back to your Radioshack system, what happens if something goes wrong? Let's say you have a problem with your high-speed SAS adapter (driver, firmware etc.). Can you call Nexenta for support, when you have a problem. Maybe they even haven't heard of the type of hba you own?"
Sun SAS HBA is a LSI rebadge. But it doesn't matter, as I have told you to buy the Sun X4440 system which is exactly what the 7410 is. So the driver, firmware...blah blah blah issues are all nonexistent.
Not that X4440 is a good choice anyways for high performance storage controller anyways. It lacks the following:
1. PCI-e 2.0
2. DDR3 ram which is cheaper and faster than DDR2 now
3. SAS 6Gbps due to lack of PCI-e 2.0
4. You are paying the 4 Socket AMD tax when the Opteron 6100s took the tax away.
Psychologically speaking, the buyer always try to compensate in their minds about how "great" a product they just purchased. I don't blame you. Enjoy your 7410, just so that you know, once you blow through your slow and tiny L2ARC, all you have is just slow ass 7200 rpm spindles below. You will hit the same metadata limitations like the NetApp you replaced. It starts around 200GB working side assuming you have 64GB ram+100GB Stec unless you are dumb enough to keep on adding the Stecs Mach8s at $50/GB.
Do you know the difference between me and you? When I tell you how to source a 7410 system from DataON and Sun X4440, you call it "Radio-Shack" gear, but you don't even know that your so called "Super doper Sun" gear is a "Radio-Shack rebadge".
"Now, if we go back to your Radioshack system, what happens if something goes wrong? Let's say you have a problem with your high-speed SAS adapter (driver, firmware etc.). Can you call Nexenta for support, when you have a problem. Maybe they even haven't heard of the type of hba you own?"
Sun SAS HBA is a LSI rebadge. But it doesn't matter, as I have told you to buy the Sun X4440 system which is exactly what the 7410 is. So the driver, firmware...blah blah blah issues are all nonexistent.
Not that X4440 is a good choice anyways for high performance storage controller anyways. It lacks the following:
1. PCI-e 2.0
2. DDR3 ram which is cheaper and faster than DDR2 now
3. SAS 6Gbps due to lack of PCI-e 2.0
4. You are paying the 4 Socket AMD tax when the Opteron 6100s took the tax away.
Psychologically speaking, the buyer always try to compensate in their minds about how "great" a product they just purchased. I don't blame you. Enjoy your 7410, just so that you know, once you blow through your slow and tiny L2ARC, all you have is just slow ass 7200 rpm spindles below. You will hit the same metadata limitations like the NetApp you replaced. It starts around 200GB working side assuming you have 64GB ram+100GB Stec unless you are dumb enough to keep on adding the Stecs Mach8s at $50/GB.
You don't get it...
Of course I know that components are sourced from somewhere, hey, even the harddisks are from Seagate and not from Sun
When I say Radioshack, I mean also, that I don't want to build it myself. What I want is an integrated and tested sytem. Not something, that "accidentally" works.
Building a Radioshack System by my own ( I could do it!) takes time (=my employers money).
If the 7410 is up to the job, for a reasonable price, it does not matter to me if it contains hamsters running in a wheel or not.
All the time you are only talking about CAPEX, which, in the end is the least costly part of it all.
If you build a system by yourself, you spend a lot of hidden OPEX (finding the components, ordering them, putting it together, testing, keeping spareparts etc.)
How much money will you save in the end?
We could change the topic to Mac vs. PC. One of them costs a little bit more, but just works out of the box...
Of course I know that components are sourced from somewhere, hey, even the harddisks are from Seagate and not from Sun
When I say Radioshack, I mean also, that I don't want to build it myself. What I want is an integrated and tested sytem. Not something, that "accidentally" works.
Building a Radioshack System by my own ( I could do it!) takes time (=my employers money).
If the 7410 is up to the job, for a reasonable price, it does not matter to me if it contains hamsters running in a wheel or not.
All the time you are only talking about CAPEX, which, in the end is the least costly part of it all.
If you build a system by yourself, you spend a lot of hidden OPEX (finding the components, ordering them, putting it together, testing, keeping spareparts etc.)
How much money will you save in the end?
We could change the topic to Mac vs. PC. One of them costs a little bit more, but just works out of the box...
Just when did I say you had to build it yourself? Heard of phone calls?
Make a phone call to Sun and say "I want X4440 with 128GB ram + Quad Opteron 8356s or 8431s the SAS HBA+ 10GigE card". Then, make a phone call to DataOn, and say, "I want a DNS1400 filled with Seagate Barracuda ES.2 1TB". Then make a phone call to Nexenta, "hey, I want a 10TB license, and have someone install it for me in 3 days."
Done.
This isn't like Mac vs PC. Because that ZFS is open source, this is more like RHEL vs CentOS.
As I said before, be happy with your 7410. People usually don't want to be informed that they have been scammed.
Make a phone call to Sun and say "I want X4440 with 128GB ram + Quad Opteron 8356s or 8431s the SAS HBA+ 10GigE card". Then, make a phone call to DataOn, and say, "I want a DNS1400 filled with Seagate Barracuda ES.2 1TB". Then make a phone call to Nexenta, "hey, I want a 10TB license, and have someone install it for me in 3 days."
Done.
This isn't like Mac vs PC. Because that ZFS is open source, this is more like RHEL vs CentOS.
As I said before, be happy with your 7410. People usually don't want to be informed that they have been scammed.
Sorry ... it isn't that easy: When i'm at a customer, i'm often doing something similar ... "you can get the all-inclusive package .. it's called S7xxx, graphical interface, validated, tuned, preconfigured, up and running in 10 minutes when you have to search for the screwdriver or you can get the kit ... buy a sun server (or whatever you want), buy some storage (or whatever you want), and use this OpenSolaris CD. You can purchase support. The problem: You have to configure and test it on your own. no gui, no analytics, but you can run whatever you want on the box additional to the server task and it's cheaper".
It's interesting that most people decide for the S7xxx. Two execeptions so far: Universities and one company, wanting to run an indexing application on the fileserver.
At the end your comment just shows again, that you are not really aware of the problems and mindset of people delivering IT services to commercial users. You don't do it on the way you proposed it, when you have a million users behind your back ....
It's interesting that most people decide for the S7xxx. Two execeptions so far: Universities and one company, wanting to run an indexing application on the fileserver.
At the end your comment just shows again, that you are not really aware of the problems and mindset of people delivering IT services to commercial users. You don't do it on the way you proposed it, when you have a million users behind your back ....
I don't understand what you're trying to say. Reading between the lines, it would appear to be: "it doesn't matter that everything you used to get from Sun will now cost lots more money because it's still a tiny percentage of your bottom line".
Nope ... i just want to say that OS, hardware and everything else is just a seconday thing on the great scale of things ....
ONLY if your customers are large corporations.
Yes, from my own experience, that is indeed so.
However, that logic is flawed, and I have evidence of that: Sun tanked and had to be sold to whoever would buy it. In the end, even IBM wouldn't touch Sun.
Without reforms, Sun is a financial black hole, not a star.
Why is the logic flawed? Because in the times of crisis, big customers consolidate, not expand. Market consolidates, not expands. Managers don't have the guts to risk their fat bonuses to expand operations, instead, they "restructure" by letting people go. (Funny thing, for all the Ferraris and Porsches and Maseratis they drive, they seem to be pretty SPINELESS when it comes to taking risks and thinking things through.)
So the big customers, which sustained these high profit margins are shrinking. The market is shrinking, dying. Daily.
The volume is in the small guy. Small and medium sized businesses. The profit lies in the selling lots and lots of cheaper-than-cheap hardware, not asking 150% for reinstatement of a support contract. That will work for another five years, then it's over. I hope Oracle milks it while it lasts, because after that, it'll be tough.
The problem is that these vendors like Oracle are too big for themselves: there is no individual responsibility, it's a free-for-all, just-make-sure-my-ass-and-bonus-are-safe kind of mentality, not the Henry Ford "I will revolutionize the car industry" kind of mentality.
Honestly, after reading "SOFTWAR: an intimate portrait of Larry Ellison and Oracle", I'm terribly dissapointed in uncle Larry, I naively believed he would put his vision from 2003 into reality. Instead, everyone is just doing the "asspalm" instead.
(TS, whoever you are, you've made history with that. So true.)
Yes, from my own experience, that is indeed so.
However, that logic is flawed, and I have evidence of that: Sun tanked and had to be sold to whoever would buy it. In the end, even IBM wouldn't touch Sun.
Without reforms, Sun is a financial black hole, not a star.
Why is the logic flawed? Because in the times of crisis, big customers consolidate, not expand. Market consolidates, not expands. Managers don't have the guts to risk their fat bonuses to expand operations, instead, they "restructure" by letting people go. (Funny thing, for all the Ferraris and Porsches and Maseratis they drive, they seem to be pretty SPINELESS when it comes to taking risks and thinking things through.)
So the big customers, which sustained these high profit margins are shrinking. The market is shrinking, dying. Daily.
The volume is in the small guy. Small and medium sized businesses. The profit lies in the selling lots and lots of cheaper-than-cheap hardware, not asking 150% for reinstatement of a support contract. That will work for another five years, then it's over. I hope Oracle milks it while it lasts, because after that, it'll be tough.
The problem is that these vendors like Oracle are too big for themselves: there is no individual responsibility, it's a free-for-all, just-make-sure-my-ass-and-bonus-are-safe kind of mentality, not the Henry Ford "I will revolutionize the car industry" kind of mentality.
Honestly, after reading "SOFTWAR: an intimate portrait of Larry Ellison and Oracle", I'm terribly dissapointed in uncle Larry, I naively believed he would put his vision from 2003 into reality. Instead, everyone is just doing the "asspalm" instead.
(TS, whoever you are, you've made history with that. So true.)
I agree with the sequence that you explain. It is obvious.
I don't agree with the money of operative system is not significant.
1- You seem to think only on big systems: SAP, etc. Most of the systems aren't of this type. Hadware/operative sytems is factor more importante in other type of systems.
2- My departament has a budget for infraestructure., and for be profitable has to reduce costs, and if my departament is expensive in its operation financial or development departaments will do a outsorcing of the infraestructure.
As Ceri Davies says, it seems that you are justifing a increase of Sun/Oracle products. It is the usual way of Oracle for make a profitable company: increase all the prices.
And you are speaking about x86 servers. I think that a Niagara Server or MX00 are a important part of the price for most of the projects.
I don't agree with the money of operative system is not significant.
1- You seem to think only on big systems: SAP, etc. Most of the systems aren't of this type. Hadware/operative sytems is factor more importante in other type of systems.
2- My departament has a budget for infraestructure., and for be profitable has to reduce costs, and if my departament is expensive in its operation financial or development departaments will do a outsorcing of the infraestructure.
As Ceri Davies says, it seems that you are justifing a increase of Sun/Oracle products. It is the usual way of Oracle for make a profitable company: increase all the prices.
And you are speaking about x86 servers. I think that a Niagara Server or MX00 are a important part of the price for most of the projects.
I'm not that sure. I just think of my brother having a small energy consulting company. He hates Windows like the plaque. Nevertheless he needs it, because of a special software to compute the fluid dynamics in piping systems. The windows license 80 Euros, hardware 1000 Euros. the software ...well... over 10.000 Euros. And that's just a two-persons consulting shop. Or special software for small businesses. Or think about the costs for Adobe creative suite compared with the operating system. Exchange a small server can handle 1000 users, dirt cheap today , but the CAL for it costs allone 67.000 Euros.
I think, the perception of software pricing is somewhat skewed by open source software downloaded for free. But many software components costs still a lot of money, as soon as you need
That said, you are right ... when you use a software that's essentially for free like a webserver or mailserver or a self developed software directly working on the , then the price for the operating system is well a factor ...
I think, the perception of software pricing is somewhat skewed by open source software downloaded for free. But many software components costs still a lot of money, as soon as you need
That said, you are right ... when you use a software that's essentially for free like a webserver or mailserver or a self developed software directly working on the , then the price for the operating system is well a factor ...
i can't believe how negative some of the comments were ... i agree with you ... but just so we are on the same page / hymn book .. this is textbook ... IT is not the business ... IT enables the business ...
there are books / volumes on this but in a nutshell ...
businesses have processes ... they are manual, but as soon as possible they get matched to an application ... depending on the scale of the businesses these applications are either customised to match the processes or the processes are changed to match the application (latter is obviously cheaper) ...
the next layer of concern to the business is the application integration layer, enterprise information integration (EII), enterprise application integration (EAI) or Extract Transform Load (ETL) ... this is how the business can mash-up the processes / application layer ... service oriented architecture (SOA) ... agility ... plug and play ... a good framework here is key to success ...
it is at these two layers (i.e. infrastructure applications) and these two layers ONLY the business realises its strategic objectives ... assuming it has a strategic plan ...
the infrastructure operations layer (i.e. networks, servers, desktops, data centres) are just tactical ... and as the author says driven by infrastructure applications ...
but it gets even more discrete ... as you read the following keep the phrase "soucing strategy" in mind ...
there are two types of applications, business support (hr, finance, legal, tax, etc) applications and business operations (crm, etc) applications ...
the business strategy is realised in the domain of business operations applications only (i.e. business processes / business operations applications / the three e's)
the business IS NOT / will not / cannot distinguish itself at the business support applications layer as the business support layer is NOT the core business ... and as such it is logical to outsource business support applications layer along with the infrastructure operations layer to focus on core competencies / core business ...
AND suprise suprise, some businesses do outsource this ...
the resulting operation is often named CORPORTATE SHARED SERVICES ...
cheers
beroccaboy
there are books / volumes on this but in a nutshell ...
businesses have processes ... they are manual, but as soon as possible they get matched to an application ... depending on the scale of the businesses these applications are either customised to match the processes or the processes are changed to match the application (latter is obviously cheaper) ...
the next layer of concern to the business is the application integration layer, enterprise information integration (EII), enterprise application integration (EAI) or Extract Transform Load (ETL) ... this is how the business can mash-up the processes / application layer ... service oriented architecture (SOA) ... agility ... plug and play ... a good framework here is key to success ...
it is at these two layers (i.e. infrastructure applications) and these two layers ONLY the business realises its strategic objectives ... assuming it has a strategic plan ...
the infrastructure operations layer (i.e. networks, servers, desktops, data centres) are just tactical ... and as the author says driven by infrastructure applications ...
but it gets even more discrete ... as you read the following keep the phrase "soucing strategy" in mind ...
there are two types of applications, business support (hr, finance, legal, tax, etc) applications and business operations (crm, etc) applications ...
the business strategy is realised in the domain of business operations applications only (i.e. business processes / business operations applications / the three e's)
the business IS NOT / will not / cannot distinguish itself at the business support applications layer as the business support layer is NOT the core business ... and as such it is logical to outsource business support applications layer along with the infrastructure operations layer to focus on core competencies / core business ...
AND suprise suprise, some businesses do outsource this ...
the resulting operation is often named CORPORTATE SHARED SERVICES ...
cheers
beroccaboy
I would be interested to know if you have worked with small and medium size businesses and if they followed the same sequence. In my limited experience, there are four main differences to the method you outline for smaller companies:
1. While a business need does start the process, it is rarely understood fully enough to provide an obvious application choice. The IT department then becomes involved early because they are thought to best understand how to map vague requirements to a technical solution. This may or may not actually be the case.
2. There are limited resources to hire additional staff with expertise in the new application infrastructure, the labor money is reserved for application development to meet the business need. Most smaller IT departments are made up of generalists and certain technologies are preferred because they already have the knowledge and toolsets to manage them in house.
3. Staff is rarely allocated solely to the evaluation and design for the new application's architecture. Hundreds of hours are not available to spend with vendors or in researching alternative solutions. The available application choices are quickly reduced to a few candidates by matching the business need with information available online and applying the technology preferences mentioned in #2. The remaining vendors are then contacted.
4. The costs of the software application and the developers to set it up take up the majority of the available budget for the project. This is where the hardware and software costs start to matter. If you can implement the solution on an infrastructure that is 50% cheaper, then that money can go to more development and integration time.
1. While a business need does start the process, it is rarely understood fully enough to provide an obvious application choice. The IT department then becomes involved early because they are thought to best understand how to map vague requirements to a technical solution. This may or may not actually be the case.
2. There are limited resources to hire additional staff with expertise in the new application infrastructure, the labor money is reserved for application development to meet the business need. Most smaller IT departments are made up of generalists and certain technologies are preferred because they already have the knowledge and toolsets to manage them in house.
3. Staff is rarely allocated solely to the evaluation and design for the new application's architecture. Hundreds of hours are not available to spend with vendors or in researching alternative solutions. The available application choices are quickly reduced to a few candidates by matching the business need with information available online and applying the technology preferences mentioned in #2. The remaining vendors are then contacted.
4. The costs of the software application and the developers to set it up take up the majority of the available budget for the project. This is where the hardware and software costs start to matter. If you can implement the solution on an infrastructure that is 50% cheaper, then that money can go to more development and integration time.
That's totally true about so called "high end commercial computing", but it seems to be a blueshifting, while redshift (i.e. fastest growth) applies to web, HPC etc, and for these markets price/performance is usually key issue, and applications are homegrown or based on open source components, and can be used in any operating environment, that is just the fastest (i.e. thay can easily move from x86 to SPARC or vice-versa if it offers performance advantages).
"A nice example is the collaboration suite: Let's assume on the shortlist of your application choice were Lotus Notes, MS Exchange and something based on open source components. When you think you can drive your user to use something to opensource, just because it's open source and just because it's portable to every platform ... just dream further. Users and especially senior users decide for something they already know. Perhaps your company works with Notes for years ... then it's highly probable that the systems will stay there .... "
This is most certainly true, from my own experience... so with that being written, why all the radical changes in OpenSolaris that broke so many different things, and made OpenSolaris unsuitable for the enterprise (case in point: no Flash(TM), no configuration packages because of IPS, etcetera)?
If WE KNOW that users dislike change and will NOT go along with them, why stubbornly proceed with breaking Solaris? Why fight the quixotic battle of trying to shove OpenSolaris down people's throats?
This is most certainly true, from my own experience... so with that being written, why all the radical changes in OpenSolaris that broke so many different things, and made OpenSolaris unsuitable for the enterprise (case in point: no Flash(TM), no configuration packages because of IPS, etcetera)?
If WE KNOW that users dislike change and will NOT go along with them, why stubbornly proceed with breaking Solaris? Why fight the quixotic battle of trying to shove OpenSolaris down people's throats?
No one is breaking Solaris. OpenSolaris is not a Solaris 10 replacement - it's an open source operating system based on the same code, that will once be used to create Solaris 11 (if it will be needed). Solaris 10 works as usual for Sun users (only problem is uncertainity about "how is Solaris 10 licensed to users of non-Sun hardware").
OpenSolaris is operating system built for different purpose and different audience, than Solaris 10. (of course if it is still alive).
OpenSolaris is operating system built for different purpose and different audience, than Solaris 10. (of course if it is still alive).
That's a very naive view of the situation, considering that Sun had stated on multiple occasions (which my employers hosted, and I attended in person), that Solaris.Next will be based on OpenSolaris.
And if, on top of that, we consider that Solaris kernel engineering at Oracle is banging their heads against concrete walls trying to solve all the architectural issues in IPS, we can conclude, with high certainty, that IPS will be the packaging "system" (and I put that in double quotes on purpose), which will be THE software management subsystem in whatever is meant to replace Solaris 10.
And that is bad. So very, very bad. A lot of customers will be royally screwed, myself included.
http://src.opensolaris.org/source/xref/pkg/gate/doc/scripting.txt
And if, on top of that, we consider that Solaris kernel engineering at Oracle is banging their heads against concrete walls trying to solve all the architectural issues in IPS, we can conclude, with high certainty, that IPS will be the packaging "system" (and I put that in double quotes on purpose), which will be THE software management subsystem in whatever is meant to replace Solaris 10.
And that is bad. So very, very bad. A lot of customers will be royally screwed, myself included.
http://src.opensolaris.org/source/xref/pkg/gate/doc/scripting.txt
I don't think that Solaris 11 will be equal to OpenSolaris.
Oracle will take the interesting parts of OpenSolaris, other parts from Solaris 10, and new developments not present in OpenSolaris for make Solaris 11. The patching/stablization of Solaris will be closed, as always, and the imported code will diverge in a fast way.
I think that OpenSolaris is crap. A system without patches, a unpredictible cycle release and broken pieces as many people indicate is a waste of time.
It is best to dedicate the time to Solaris or other UNIX like CentOS/Debian/XBSD
It is only my thinking. Oracle isn't showing any important commitment with OpenSolaris. It is a toy for do new developments in Operative System. It has the plus that a few people work free for Oracle adding/fixing things for OpenSolaris and serves as flag of OpenSource for marketing.
Oracle will take the interesting parts of OpenSolaris, other parts from Solaris 10, and new developments not present in OpenSolaris for make Solaris 11. The patching/stablization of Solaris will be closed, as always, and the imported code will diverge in a fast way.
I think that OpenSolaris is crap. A system without patches, a unpredictible cycle release and broken pieces as many people indicate is a waste of time.
It is best to dedicate the time to Solaris or other UNIX like CentOS/Debian/XBSD
It is only my thinking. Oracle isn't showing any important commitment with OpenSolaris. It is a toy for do new developments in Operative System. It has the plus that a few people work free for Oracle adding/fixing things for OpenSolaris and serves as flag of OpenSource for marketing.
"I can't count the number of occurances, where a customer said to me "cool system, but it isn't in our book of standards" or "cool system, but we have already an agreement with vendor X/Y/Z for this class of system"."
Yeeeepppp!!!
Hi Joerg! These are the places I used to work at (:-)
Ah the bad old days... I don't miss them at all!
Nothing wrong with a standard set of hardware -- it's extremely cheap for the customer -- but the "standard set" of approved hardware often needs a revision to the choices and options.
There were a lot of politics: "we have to go with SPARC. Management will never approve going with x86".
Boy am I glad all of that is just a bad, bad memory now.
Yeeeepppp!!!
Hi Joerg! These are the places I used to work at (:-)
Ah the bad old days... I don't miss them at all!
Nothing wrong with a standard set of hardware -- it's extremely cheap for the customer -- but the "standard set" of approved hardware often needs a revision to the choices and options.
There were a lot of politics: "we have to go with SPARC. Management will never approve going with x86".
Boy am I glad all of that is just a bad, bad memory now.
"Oracle DB isn't an application. It's middleware. A SQL database alone solves no business problem."
Ho ho ho!
That's a very bold statement, considering I could write, and have maintained, entire credit contract reconciliation APPLICATIONS which ran INSIDE of the Oracle kernel!
The entire app was written in PL/SQL. It ran solely in and of itself in Oracle RDBMS.
Ho ho ho!
That's a very bold statement, considering I could write, and have maintained, entire credit contract reconciliation APPLICATIONS which ran INSIDE of the Oracle kernel!
The entire app was written in PL/SQL. It ran solely in and of itself in Oracle RDBMS.
Yes ... but was the application you've written that solved the business problem, not the Oracle DB. The OracleDB just provided the infrastructure by providing the database and the runtime environment for your PL/SQL code. A Oracle DB just do nothing without having some code querying the data.
Joerg!
How about keeping it fair? Fair!
What you are doing right now, is trying to bust me on semantics. Life isn't fair, and in this particular case, neither are you.
We do not need to discuss solving a business need. I myself am a very, very firm believer that computers exist to solve problems, not create new ones. So a computer is there to DO WORK.
All I did was show you, that Oracle is not "middleware", and that it can easily be, and for employers I worked for IS the application.
Here is the deal: when I am wrong, I've got the GUTS to admit I am wrong. Publicly, if necessary. We talked about having guts before.
Is it really too much to admit you were wrong, at least on that count? Really?
How about keeping it fair? Fair!
What you are doing right now, is trying to bust me on semantics. Life isn't fair, and in this particular case, neither are you.
We do not need to discuss solving a business need. I myself am a very, very firm believer that computers exist to solve problems, not create new ones. So a computer is there to DO WORK.
All I did was show you, that Oracle is not "middleware", and that it can easily be, and for employers I worked for IS the application.
Here is the deal: when I am wrong, I've got the GUTS to admit I am wrong. Publicly, if necessary. We talked about having guts before.
Is it really too much to admit you were wrong, at least on that count? Really?
1. I hope you have no problems, that i moved your comment at a different position in the thread, as it looked to me as an answer to 9.1 ...
2. However: You are vastly undervaluing your work by calling the Oracle RDBMS the business-problem solving entity. It was your work, not the one of an unnamed Oracle programmer. This would be the same like calling the php binary an application or the Java Application Server an application. php is nothing without the code written solving the business problem programmer, the Java App Server is nothing without the Java code developed by an programmer.
This isn't about semantics from my point of view. It's pretty important that an App Server, all the middleware products, all the databases do nothing without the work of a programmer writing the code to solve the business problem. It wasn't the Oracle RDBMS that did the credit reconciliation, it was the code that you've written, using the database as it's executing environment.
3. There is indeed an error in my statement: A database management system isn't middleware, in most architecture textbooks it's an own layer in the overarching image of an IT systems architecture.
2. However: You are vastly undervaluing your work by calling the Oracle RDBMS the business-problem solving entity. It was your work, not the one of an unnamed Oracle programmer. This would be the same like calling the php binary an application or the Java Application Server an application. php is nothing without the code written solving the business problem programmer, the Java App Server is nothing without the Java code developed by an programmer.
This isn't about semantics from my point of view. It's pretty important that an App Server, all the middleware products, all the databases do nothing without the work of a programmer writing the code to solve the business problem. It wasn't the Oracle RDBMS that did the credit reconciliation, it was the code that you've written, using the database as it's executing environment.
3. There is indeed an error in my statement: A database management system isn't middleware, in most architecture textbooks it's an own layer in the overarching image of an IT systems architecture.
"I hope you have no problems, that i moved your comment at a different position in the thread, as it looked to me as an answer to 9.1 ..."
No, no problem at all, by all means. That must've been an error or oversight on my part. Thank you kindly for that correction.
"There is indeed an error in my statement: A database management system isn't middleware, in most architecture textbooks it's an own layer in the overarching image of an IT systems architecture."
Actually, in most cases, an RDBMS is a storage engine with a consistent retrieval mechanism (SQL).
I guess this is as close as I'm ever going to get to you admitting a mistake, so fine, I'll take what I can get in this particular case. At least, it's progress.
No, no problem at all, by all means. That must've been an error or oversight on my part. Thank you kindly for that correction.
"There is indeed an error in my statement: A database management system isn't middleware, in most architecture textbooks it's an own layer in the overarching image of an IT systems architecture."
Actually, in most cases, an RDBMS is a storage engine with a consistent retrieval mechanism (SQL).
I guess this is as close as I'm ever going to get to you admitting a mistake, so fine, I'll take what I can get in this particular case. At least, it's progress.
Whenever i see i've made an error, i correct that error ... happened in the past ... i've made even such a big error once (about Sun in India) that resulted into an own article for the correction with an apology ...
But i'm just doing that, when i made an mistake ... and in this situation i just don't think my statement was incorrect, that Oracle isn't an application.
By the way: When i'm looking at all the discussions with you and you consider me of having a problem of admitting an mistake, i think you have the same problem of admitting mistakes
However: You have your opinion, i have my opinion, i'm explaining my opinion, you are explaining your opinion .... it's about discussion, not convincing at all costs. When you need me to admit an mistake, fine ... will do it the next time i see one ...
However you don't have to admit it when you make one ... i'm not really interested in this kind of satisfaction. I'm interested in a constructive discussion worth the time i'm spending for it. I've learned in the past that there is a multitude of perspectives, and each one of them has it's validity. I learn a thing or two with each discussion and i hope others learn a thing or two from my articles or comment or at least respect my opinion. What i'm hating like the plaque are fruitless, intellectual uninteresting discussions or discussion participants stating opinions again and again that conflict in an obvious manner with commonly available empirical data or knowledge or resulting from a egocentric and arrogant perspective to the world.
Of course there is some satisfaction in hearing that someone says that she or he thinks i'm right, but that satisfaction comes from the transfer of information, of knowledge. Discussion is not a sport counting convinced vs. being convinced. That the reason, why a mail stating that my lksf helped someone to understand solaris is infinitely more important to me, than someone stating that he made an mistake.
That said, i like a good, intellectual interesting and challenging discussion ... so i don't go out of the way of them ....
But i'm just doing that, when i made an mistake ... and in this situation i just don't think my statement was incorrect, that Oracle isn't an application.
By the way: When i'm looking at all the discussions with you and you consider me of having a problem of admitting an mistake, i think you have the same problem of admitting mistakes
However: You have your opinion, i have my opinion, i'm explaining my opinion, you are explaining your opinion .... it's about discussion, not convincing at all costs. When you need me to admit an mistake, fine ... will do it the next time i see one ...
However you don't have to admit it when you make one ... i'm not really interested in this kind of satisfaction. I'm interested in a constructive discussion worth the time i'm spending for it. I've learned in the past that there is a multitude of perspectives, and each one of them has it's validity. I learn a thing or two with each discussion and i hope others learn a thing or two from my articles or comment or at least respect my opinion. What i'm hating like the plaque are fruitless, intellectual uninteresting discussions or discussion participants stating opinions again and again that conflict in an obvious manner with commonly available empirical data or knowledge or resulting from a egocentric and arrogant perspective to the world.
Of course there is some satisfaction in hearing that someone says that she or he thinks i'm right, but that satisfaction comes from the transfer of information, of knowledge. Discussion is not a sport counting convinced vs. being convinced. That the reason, why a mail stating that my lksf helped someone to understand solaris is infinitely more important to me, than someone stating that he made an mistake.
That said, i like a good, intellectual interesting and challenging discussion ... so i don't go out of the way of them ....
Links in this article
The LKSF book
The book with the consolidated Less known Solaris Tutorials is available for download here
Web 2.0
Contact
Networking
xing.com
LinkedIn
Facebook
My photos
- mail: joerg@c0t0d0s0.org
- icq: 144000169
- Me, myself and I
- openBC
- My del.icio.us
- c0t0d0s0@twitter
- c0t0d0s0@last.fm
- wiki@c0t0d0s0
- Amazon Wishlist
Networking
xing.com
My photos
Syndication
Tagged articles
AMD Apple avs Bahn Blogging Blogosphere braindump Business Travel CeBIT cec cec2006 CMT del.icio.us deutsch dtrace fliegen Fundsache General Hamburg IBM i hate sundays Intel iscsi jumpstart Links Linux lksf Mindfuck Movies Music Musik Niagara Opensolaris Opteron Photographie policy of ... Politik Security Solaris storage Sun suncec2007 sunw t1 The IT Business Ultrasparc ultrasparc t1 Wirtschaft Work ZFS
Comments
about links for 2010-09-05
Mon, 06.09.2010 08:52
besides pissing everyone of in
the company hurd is most prom
inent for slashing r&d at hp t
o its bare minimum. famo [...]
about A really long rant ...
Mon, 06.09.2010 08:19
>termination. I don't think th
e ON gate >was ever showing re
al time >development, just the
staged >releases of cod [...]
about links for 2010-09-05
Mon, 06.09.2010 03:34
Close OpenSolaris, ramp up sup
port for Solaris 10, sue Googl
e over Java, let star engineer
ing talent go - and now [...]
about A really long rant ...
Sun, 05.09.2010 21:03
Looks like I'm a bit late to t
he party here, so maybe no-one
will read this, but here goes
anyway:
First, in my o [...]
about A really long rant ...
Sun, 05.09.2010 19:02
I know what you mean with your
talk about packaging.
And th
at's what I was saying: you ca
n package yourself your [...]
Buttons
This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Germany License
