The Sun Blade Blog
Would IBM "Dare To Compare" blade offerings more accurately? (#1)
On its website, IBM published a competitive analysis of its
BladeCenter offerings versus blade products from HP, Dell, and Sun. Of
course, assessments on Sun's blade products were negative,
misrepresented, and in some areas outright wrong, while positioning
IBM in a positive light. The IBM blade "Dare To Compare" web page is
posted
at:
http://www-03.ibm.com/systems/migratetoibm/systems/bladecenter/compare2.html
Sun has made IBM aware of their factual errors about the Sun Blade
products and IBM will eventually correct their blade "Dare To Compare"
web page. In the meantime, to make things clearer here is one claim
where IBM is wrong about Sun Blades. We will share a few more of these IBM 'misrepresented truths' over the next few days ... so stay tuned.
IBM FUD: IBM BladeCenter H offers better chassis density and
BladeCenter E offers better rack density
Sun Truth: IBM is
misleading and only comparing
the density number
of server blades in a chassis and rack. They neglect to
incorporate other important aspects of number of processor sockets and
cores, memory, I/O and power. In reality, when
considering
all aspects of a blade offering - balance of CPU, memory, I/O and power
-
Sun offers better blade server flexibility than IBM. The BladeCenter H and E
chassis', with 14 single-wide HS21 quad core Xeon server blades,
forces a customer to trade the number of server blades against the
cost of losing processor socket, memory and I/O density in the
chassis and rack; as well as possibly over subscribing chassis power
or running blade workloads at less performance. In
addition, IBM neglects to include the Sun Blade 6048 in their
comparison claims. The Sun Blade 6048 offers comparable CPU
socket density per chassis and exceeds the IBM BladeCenter offerings up
to 1.71X in
terms of supported overall CPU core density per rack (768 vs 448), with 2X the
memory DIMMS and
capacity per blade (8 DIMMS vs 16 DIMMS, 64GB vs 128GB), and up to 2.5X
the I/O bandwidth per blade (110 Gbps vs 40 Gbps).
Posted at 12:41PM Mar 05, 2008 by Douglas Wilson in Sun | Comments[7]
Who exactly is misleading? We're a Sun Blade 6000 customer, but I'll call the same sort of "BS" as you have on the IBM posting. Your comparison of "core density" only comes to play if you're talking Niagra platform. I'll be plain about this - SPARC is irrelevant in the blade space, except in the niche markets where SPARC customers are interested in blades, but haven't yet jumped ship to the X86/X64 realm (a smart move on Sun's part to help keep that momentum). Sun wins handily on RAM density (we dig our 32GB ESX blades) and I/O, truly. But by certain metrics, their kit is better. Even Dell's new offering has some plusses.
As with all technology decisions, the real answer to "which is better" is "it depends." The question is better asked "which is better for my needs." In our case, the 6000 was the right choice. We're soon going to take care of all that I/O capability by deploying our first ever 10Gbps hosts. Standard PCIe ports was another huge plus favoring Sun in our decision.
Posted by Charles Soto on March 05, 2008 at 04:34 PM PST #
Charles,
Just curious about why those PCIe slots were important to you. It seems like all the major blade vendors offer the ability to use a PCI expander blade to run one-off customer PCI, PCIe cards (which would not be a good solution if every server needed one of these cards for sure). I'm just curious what drove that decision for you. Also, did you evaluate HP or IBM blades before you bought the SUN blades? I agree that the new Dell blades (M1000e) look "interesting". I need to see one up close though to see how good (or poor) of a job they did copying HP.
Thanks
Posted by Jeff Allen on March 06, 2008 at 09:48 AM PST #
Charles, thank you for your comment on this blog posting. It was not my intention to mislead you in regards to core density on Sun Blades. In reality when I stated support for 'up to' 768 cores in a rack, that was with the Sun Blade 6048 and indeed with X86/X64 blades - not the Niagara blade that you referred to. With upcoming X86/X64 quad socket quad core blades, you can get 768 cores in a Sun Blade 6048 chassis, which is 1.71x more cores than IBM can get in a rack of Bladecenter H and HS21 Xeon blades. For an example of how this Sun Blade 6048 density is being used for customers, take a look at http://www.taborresearch.com/tacc.html for the Texas Advanced Computing Center. And yes, there's physically 768 X64 cores in each of the SB6048 systems deployed.
Posted by Doug Wilson on March 06, 2008 at 12:29 PM PST #
Hi,
What do you guys thing of the following comparison between IBM & HP Blades http://itcomparison.com/Blades/Bladesibmvshp/Bladesibmvshp.htm
It seems the site is quite open about their comparison as well and requested to post opinion and comments about their comparison at http://itcomparison.com/phpBB2/viewtopic.php?t=47
I was wondering how true is that comparison & if there is any points they had mislead it that we can point on their forum and see how serious they are about fixing it.
Thanks,
VMguru007
Posted by vmguru007 on April 03, 2008 at 10:10 PM PDT #
Hello VMguru007, thank you for sharing the links to the itcomparison.com site and asking for my opinion. In my opinion these are very high level comparisons and not in detail. It even seems that the supposed 'unbiased' comparison only took IBM's own Dare To Compare Blade web content and reiterated IBM's own claims. Thus I can't see this as an unbiased comparison with actual analysis work done. In regards to your question about where IBM has mislead in its comparisons - well, 4 of them are posted here at (http://blogs.sun.com/blades). There were more. All of which were directly communicated to IBM for fixing on their web site. I have to assume that HP has contacted IBM on the same misleading claims against their blade products.
Posted by Doug Wilson on April 14, 2008 at 11:01 AM PDT #
Hi Doug,
I am not sure how you came up with the comparison point by vmguru007 is being biased toward IBM where that site is even showing many blades comparisons where IBM is not even a part of it. An example of that:
Dell vs HP
http://itcomparison.com/Blades/Bladesdellvshp/Bladeshpvsdell.htm
what do u think further of HP vs SUN? Is this one biased toward HP !!!!!
http://itcomparison.com/Blades/Bladeshpvssun/Bladeshpvssun.htm
Is that site only created to bias against SUN?
Hmm no hard feeling I hope.
Enjoy,
Virtualization Master
Posted by Virtualization Master on December 17, 2008 at 10:03 AM PST #
One more point for IBM a new 4 Socket Intel Blade is coming up :). Check out the post at: http://www.tsmguru.com/blades/ibm-blades/ibm-planning-to-release-four-socket-intel-blades.html
With this IBM will have every required configuration :).
Posted by TSMGuru on September 23, 2009 at 11:26 AM PDT #