c0t0d0s0.org

Over a year ago, we have announced the Sun Modular Datacenter 20. Some people asked about our mental status because of this announcement. But it doesn´t seem so dumb at all, as HP has announced a datacenter container on it´s own: HP Performance-Optimized Datacenter

Many people think about blades as the best way to put many small servers in a small amount of space. Often this is the case, but not in every case. Last weekend i did some research for a project and found out something interesting in the offering of our competitor HP.

Let´s assume you have the requirement to use a quad processor blade. So you need one of the full height blades of HP the two disks of the blade aren´t sufficent your you. You need more of them. Our beloved competitor has a blade shelf for this. It´s called HP StorageWorks SB40c storage blade. It´s a half height blade.

Okay, let´s calculate a little bit. You could put 16 half height blades and 8 full heigt blades into a 10 rack units c7000 blade enclosure. Okay, we need some bays for the storage. You would assume that you can put 5 full height blades into the system and 5 storage blades in the free remaning bays (leaving one free) into the system, but this isn´t the case. The storage blades have to be adjactent to the blade. More important, you can put it only in the lower bay.

A half-height SB40c storage blade must be installed within the same partition as its partner server
blade. If the SB40c is partnered with a full-height server blade, the SB40c must be installed in the
lower bay adjacent to the full-height server blade. The HP BladeSystem Enclosure Tech Brief states:

You could put a server into the free bay above the storage bay but there is an interesting gem in the documentation. The quick spec document states at the HP website states:

NOTE: The lower tape or storage blade cannot be removed without first removing the upper half height blade.

The problem, as far as i understand the datasheet of the SB40c it isn´t just a bunch of cabling. It has an own RAID-Controller( HP Smart Array P400 with 256MB Battery Backed-Write Cache). It´s an active component and there is some likelihood that active components will fail, especially when they contain batteries.

Let´s assume you have used the slot above the Storageblade for another system. Let´s assume you have to service the storage blade. You have to shut down the perfectly running system on the blade above the storage blade, as you have to remove it first. Doesn´t sound reasonable for enterprise usage. This leads me to the conclusion that you you use the c7000 enclosure only for 4 full height quad socket blades with 4 storage blades and nothing else.

Okay, that implies that you need 10 rack units (the size of the c7000 enclosure) to implement 4 quad socket systems with 3-8 disks with the HP blades. Now let´s look at the problem from a different perspective: The Sun Fire X4450 is a quad socket socket server with up to 8 harddisks (with more PCIe-slots than a blade). The system takes rack units in you rack. So you could place 5 of them at 10 Rack units. Sometimes the denser solution is at a place where you don´t expect it.

Sometimes i have the strong impression that Linux is the landfill for old filesystems. Linux is now the proud home of AdvFS. as HP contributed the old Tru64 Unix file system to Linux. But: Does Linux really need yet another filesystem? Beside all the other ones ?

In case you don´t want to use Sun x86 servers, there is a good reason to use Solaris on it: Solaris Cluster is supported on HP Proliant DL 385 G2 and Proliant DL 585 G2 server. More details on the Sun Cluster Open Hardware page .

In IBM Systems Magazine you will find a comparision about the virtualisation technologies from IBM, HAP and Sun. Okay, you won´t expect an unbiased comparision. But this article needs some corrections. Two of colleagues done the job to read through this article and to revisist the statements. Response to: IBM, Sun and HP: Comparing UNIX Virtualization Offerings by Jeffrey Savit or a identical titeled article by Jim Laurent.

You can say everything about HP, for example that they are an printer supply company with an descending server business or that they selled the business unit with the best hp products, when they´ve spinned-off Agilent or that they are bound to an increasingly insignificant itanic processor technology, that will have it´s breakthrough "real soon now" (as in the last few years) , but you can´t say that their messages to the press are not highly entertaining: HP supports Solaris 10 on their Servers and packages this message in HP Simplifies Migration for Sun Solaris Customers to More Affordable, Higher-Performing HP ProLiant Systems"

Well, the one billion are the perhaps the number of old E250 that gone out of service, when we hat no descent opteron server in our portfolio. Dear HP, i would find the numbers quite interesing, how your Opteron/Xeon business predated your high-margin PA-RISC business. And your HPUX is years behind Solaris, and heck ... is there an HPUX for Opteron ?

The good message after all: HP supports Solaris 10, so i welcome you to be one of our valued channels for Solaris adoption.

Einen habe ich noch zum Thema HP-UX. Paul Murphy schreibt in seinem Blog:

Oddly, I think irt does, because if you're still using HP-UX there's an important message here: it's way past time to move to either Linux or Solaris, because your preferred vendor is busily petting you on the head with one hand, while doing nothing for you with the other.

Das kommt halt davon, wenn man sein Unix bei einem Unternehmen für Druckerbedarf kauft

Das entwickelt sich zu einer äusserst unschönen Angelegenheit für HP: Just when you thought HP's PatriciaGate couldn't get worse, it does. Es scheinen dort wohl Keylogger zum Einsatz gekommen zu sein. Ich bin gespannt, wie es da weitergeht ....

It´s interesting how the times have changed. Not long ago, Sun carried the meme "Good hardware, but expensive" with it. Now the Inquirer reports, that we are cheaper than HP and Dell.

I have a good rule of thumb: The more bullshit someone tells about you, the more fear he has of you. And HP seems to be in desperate fear according to this FUD-Mail from HP. Just one note: There were no customer interest for 8 Socket server because there were no 8 Socket server from a Tier1-Supplier. And the few existing ones are not exactly well engineered. Let´s wait. The x86 world changes this evening a little bit more in Sun´s favour..

No, a fan from HP. I´ve already reported about the new fan technology out of the HP labs. This is what you get, when you solve the cpu heat problem at wrong end. HP designed this beast of a fan.

It´s obvious you can find better solution. You can buy already one from Sun. The solution is called UltraSPARC T1

(via cuddletech and uadmin)

Sorry, but this sounds to me like moving deck chairs on the Titanic. The solution of the heat problems on x86-Systems is not reducing the power consumption. The solution are better fans. Wow. What a ingenious idea. And what do you make with the hot air in the datacenter. Even faster fans in the climate control systems ? Hot air doesn't disappear by blowing it out of your server. And this is the rationale behind liquid in the datacenter: You need it in your server to get the heat out of you racks and to avoid hot spots in your datacenter layout.
So the argument "I think I just found a way to keep liquid out of our servers for another 10 years." is utter bullshit, the wording "I think I Just found a way to keep our blades from wrecking up by heat until the next intel heatplates are available" would be much better. Don't take me wrong: It's a nice solution to make better fans, but it doesn't really solve any energy problem. Sun developed UltraSPARC T1 exactly out of this reason. You don't have cool down air, you havn't heated up at the first place.
But this is the problem, when you have almost no processor technology you develop on your own. You can't solve the root cause, so you have to move the deck chairs. And hell, the rails at the bow needs some paint from all the idiots standing there and shouting "I'm the king of the world".

What is the rationale behind a connector that only used in systems with a few processors? Surely, HTX is an interesting technology for systems where you really need an really-badass-ultralowlatency interconnect. But to be realistic: the only market for this seems to be HPC. This is a relativly small market , and even in theis market there there are only a small amount of customers who need more than the performance offered by PCI-Express.
Further problems that came in to my mind : Will there be a ecosystem of card manufacturers? Will be there more cards than a few Infiniband cards ? Are the cards really capable to use the advantages or do they use the same chips only with a different interface ? Will the market big enough to justify the development of adapters that take full advantage of HTX.
You must take all this factors into consideration when you sacrifice one of your slots. For a Tier-1-manufacturer its more important to put more FC- or networkcontroller with PCI-E or PCI-X into their systems than a special purpose connector for a special protocol. And with this arguments it makes sense, not to use HTX.
But i'm sure that IBM and HP will sacrify a port only to have a bullet more in their data sheet and i`m sure there will be enough customers who use this bullet to explain their decision against Sun with this bullet point. But at the end we will have a vast amount of unpopulated HTX-Ports. And at last: There has to be a niche for Tier-2 and Tier-3 companies.

Scott writes an open letter to the CEO of HP: Let's converging Solaris and HP-UX. At first most people will think: What a ridiculous idea. But think twice. Let's assume that Linux is not the answer to all question in enterprise computing. Let's assume that you are are the manufacturer of a server system with a realatively small footprint in the market. You have immense costs do develop a competitive derivate of UNIX for your own plattform. So it can make sense to partner with others to use a common commercial unix. With advantages for all sides. The ISVs have a common unix environment for all platforms and have only to recompile their code. The end users have only to establish a single version of their runbooks and operation procedures. And the manufacturers can develop jointly one commercial grade unix with all their experiences of years on unix development.
When you really think about it, it's an offer HP can't refuse. Even when you are called a pig with make-up

There is a meme in IT: Linux is cheap. Even smart people like Andreessen are surprised, that this meme proofs more and more as false:

As Ning figured such things, the loaded cost (AKA TCO) of a whitebox Linux box running Intel chips was $10,350 over 36 months. Sun's AMD boxes running Solaris 10? $4760. Whatever you might think about those numbers and how they were generated, it was pretty apparent that they surprised the Ning gang.

Okay, more news: TheRegister reports about our new workstations. VMWare seems to feel the heat of the competition. There are rumors of a free version of GSX.
The effect of "Frankenstein Computing" gets more visible from time time:

"'Best of breed' is now code for Frankenstein. I've walked into more than a few data centers where I said, 'Whoa, where did you get that thing?'" McNealy said. "They've got body parts from every supplier you can imagine. There are big bolts sticking out everywhere and stitch marks."

.
But in most cases its not the fault of the IT-staff. Nowadays the decisionmakers do not decide with techical expertise, they collect numbers for theirs spreadsheet. Most of the time decisionmakers have no technical background, they have a major degree in bean counting. In conjunction with clueless project managers the worst combination, you can imagine. In this situation the only valid solution is to leave the table to rescue the remaining parts of your own sanity. At the end the customer has a datacenter consisting out of patchwork technology, nothing really matches together and he is constantly occupied with plugging eleven holes in the dam with ten fingers. Okay, enough lamentation ...
The newest fashion fad in frankenstein computing: Watercooling in the Datacenter. Okay, we are not able to drive down heat disipation, let´s use water like 20 years ago. A old kludge to solve a problem with a root cause: Using a desktop processor as a server processor. But , dear HP, i see the problem. This the only you can solve it, when you outsource all your processor development to Intel and AMD. But to be honest, in my opinion water cooling is the only way to increase the density of x86-computing. The question is: Is x86 - at least with the current technology - the correct solution at all? And what is the right solution? Maybe the industrie has to start again, because even the best kludge cannot last forever.
A few days ago i read a nice quote in a internal mail . Basically it goes like: The biggest intellectual capital is the courage to build something like Niagara And with this words i close for today.