Quicksearch
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.
Sun SPARC Enterprise M3000 announced
Tuesday, October 28. 2008
Sun and Fujitsu jointly announced the system codenamed Ikkaku today : Sun SPARC Enterprise M3000.
The M3000 is a single-socket system for the SPARC 64 VII CPU, thus itīs a 4 core system, executing up to 8 threads. Up to 32 GB memory. 4 PCI-Slots. So itīs an entry-level system. Starting at $14.795,00.
But before you compare it with your favourite brand of el-cheapo x86 systems. This system shares many RAS features of his bigger brothers: Instruction Retry on processor level ... ECC for memory ... ECC for the interconnect between system controller and the CPU ... ECC for integer registers ... parity for floatingpoint registers ... fault isolation ... a single defective core doesnīt keep the system from starting up, etc. ... well ... just read the architecture whitepaper on your own.
The M3000 is a single-socket system for the SPARC 64 VII CPU, thus itīs a 4 core system, executing up to 8 threads. Up to 32 GB memory. 4 PCI-Slots. So itīs an entry-level system. Starting at $14.795,00.
But before you compare it with your favourite brand of el-cheapo x86 systems. This system shares many RAS features of his bigger brothers: Instruction Retry on processor level ... ECC for memory ... ECC for the interconnect between system controller and the CPU ... ECC for integer registers ... parity for floatingpoint registers ... fault isolation ... a single defective core doesnīt keep the system from starting up, etc. ... well ... just read the architecture whitepaper on your own.
Trackbacks
Trackback specific URI for this entry
No Trackbacks
Comments
Display comments as
(Linear | Threaded)
Looks like the long missing replacement for V245 and the likes.
Yes, together with the T5120/40 depending on the job ...
Bit worried about the I/O - none of the pdf put figures to any of the buses.
On the point of el-cheapo x86, you can buy 7 for that price.
Also ram for it is not that special, so why charge that much for it?
On the point of el-cheapo x86, you can buy 7 for that price.
Also ram for it is not that special, so why charge that much for it?
1. What number regarding the I/O do you want to know?
2. There are many jobs where the six additional servers doesnīt help you to get a better availability. And think about all the other RAS ... this gives you a notable more stable system.
2. There are many jobs where the six additional servers doesnīt help you to get a better availability. And think about all the other RAS ... this gives you a notable more stable system.
Any joy with the I/O numbers?
From the look the PCI-E bus only has 16 channel - x8 for each switch.
So it looks a little starved on the I/O front.
From the look the PCI-E bus only has 16 channel - x8 for each switch.
So it looks a little starved on the I/O front.
4 GByte/s. So you are correct. But i don't think about it as starved. You have to source and sink that on the CPU.
So you don't count nics and sas as I/O then?
Start using the 4 pci-e for something like 10g nics and sas hba and things get crowed on the switchs.
Any idea when the docs are going up on docs.sun.com?
Start using the 4 pci-e for something like 10g nics and sas hba and things get crowed on the switchs.
Any idea when the docs are going up on docs.sun.com?
Of course you can construct a scenario, where the I/O bandwidth of the bridges doesnīt suffice but letīs look a little bit more realistic to it. Do you really think you will fully load all interfaces at top speed at the same time? While doing something usefull with the data with 8 threads?
As for the seven server thing:
Of course you can buy six additional servers for the money.
But: Do you really want to?
It's then seven servers to burn money on to powern on, to cool, to install and maintain.
Six more servers mean also six more support contracsts and so on.
If you put in the costs for licences like Veritas or Oracle or anything like that it's pretty meaningless how much the acutal hardware costs -- you will probably pay much more on software.
I'm not saying that x86 Servers especially from HP or SUN are bad, quite on the contrary, but just complaining about the one-time investment for the hardware is pointless.
Of course you can buy six additional servers for the money.
But: Do you really want to?
It's then seven servers to burn money on to powern on, to cool, to install and maintain.
Six more servers mean also six more support contracsts and so on.
If you put in the costs for licences like Veritas or Oracle or anything like that it's pretty meaningless how much the acutal hardware costs -- you will probably pay much more on software.
I'm not saying that x86 Servers especially from HP or SUN are bad, quite on the contrary, but just complaining about the one-time investment for the hardware is pointless.
Well no.
You could just buy two for about 2/7 of the price of the m3000 and cluster - like everybody else.
Not sure about power - sparc64 in it pulls 135watts - so you could splash out and get the low powered versions and be within the m3000 power budget.
So you can get close to m3000 availability for a fraction of the price - and it will be cheaper when the i7 comes out.
You could just buy two for about 2/7 of the price of the m3000 and cluster - like everybody else.
Not sure about power - sparc64 in it pulls 135watts - so you could splash out and get the low powered versions and be within the m3000 power budget.
So you can get close to m3000 availability for a fraction of the price - and it will be cheaper when the i7 comes out.
The answer to to availability is not always "buy two of them and a cluster". The availability may be numerical similar on the OS level, but as soon as you take the application into consideration, itīs a different story.
Letīs assume, you have a long running analysis job on a machine. Of course you could restart it on a second machine, but your work is lost. So you want a machine thatīs as stable as possible.
Or Oracle. A failover from one system to another takes some time (up to minutes). Of cource you could invest in Oracle RAC as an failover accelerator (the only usecase in my opinion) but that costs you 47.000 Euro. Sometimes itīs cheaper to buy a system, thatīs more available in its self.
And as Joern stated correctly. Initial Investment costs are pointless. You have calculate all the costs over three years into a system. And then the story looks differently. But just donīt make the assumption that any x86 server is as available as a SPARC server.
Letīs assume, you have a long running analysis job on a machine. Of course you could restart it on a second machine, but your work is lost. So you want a machine thatīs as stable as possible.
Or Oracle. A failover from one system to another takes some time (up to minutes). Of cource you could invest in Oracle RAC as an failover accelerator (the only usecase in my opinion) but that costs you 47.000 Euro. Sometimes itīs cheaper to buy a system, thatīs more available in its self.
And as Joern stated correctly. Initial Investment costs are pointless. You have calculate all the costs over three years into a system. And then the story looks differently. But just donīt make the assumption that any x86 server is as available as a SPARC server.
By the way ... the price for the memory is largely based on extreme quality standards.
Its the fsb - link to the pci-e bridge - how many pci-e channels do you get.
From the block diagram it looks like a fudge again with the switchs - are the switchs any good? Any benchmarks?
From the block diagram it looks like a fudge again with the switchs - are the switchs any good? Any benchmarks?
+1
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
Comments
about Nanosecond
Wed, 23.05.2012 00:11
I remember this being drummed
into us during Digital Design
at Uni. It's important to cons
ider it when laying out [...]
Mon, 21.05.2012 18:04
Hello Kevin, Im not surprised
with what you are seeing or ha
ve seen when attaching a SSD t
o a USB2.0. USB3.0 helps [...]
Mon, 21.05.2012 04:44
Hi Greg,
With regards to IO
PS I have seen terrible result
s using a 60GB SATA2 SSD with
USB2.0 - USB2 really cho [...]
about ZFS Dedup Internals
Sat, 19.05.2012 09:50
There is no impact to boot/imp
ort times, as the DDT is loade
d as needed ... so the pool is
imported as fast as wit [...]
Buttons
This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Germany License
