Server racket to slow in 2009? With some 2 million-plus physical boxes still being shipped every quarter on a global basis, the server racket is still a pretty good one. In the first three quarters of 2008, vendors peddled some $40bn in gear, and if you use IDC data as a gauge, they've collectively managed to boost revenues by 3.5 per cent for the year so far …
Expect IBM and HP to gain share then really grow when the economy recovers.
Expect Sun to become like all the other IT has beens: Unisys, DEC, DataGeneral, Burroughs, Amdahl, Sequent, Cray, SGI, Honeywell.
While some are still somewhat alive they are all ilrelevant.
Dell? The low cost of acquisition model is not as important as TCO with everything going virtualized they are out in the cold.
Pirate hardy har har
It's interesting to see how Supermicro and Verari seem to do so well in these tough times. Maybe IBM, HP, Dell or Sun should consider buying them and trying to own the super-data-center build-outs. I wonder if Rackable will survive 2009? I wonder if Dell will buy Sun? Hmmm interesting times...
I disagree your statement that a 4 socket x86 server costs $100k. I just checked both dell.com and hp.com and could configure a 4 socket (6-cores per CPU) server with various configurations ranging from 16GB to 128GB. The cost of HW plus basic warranty was between $13k to $22k.
I do agree with your comment that 2 socket servers may be sufficient for most jobs. This has been the case for many years and will likely be for many to come and is validated by any of the major analyst reports, IDC, Gartner, Mercury. These servers offer sufficient performance for the majority of workloads that exist, vitrualized or not virtualized, and offer exceptional performance per $ and performance per Watt.
However, a 4 socket server has it's place as a purpose built server and are often deployed by IT for a certain class of business applications (database, ERP, svr consolidation) where more compute, memory, I/O and reliability are required to handle the task. I would never buy a Hummer on MPG. I don't think this fundamentally changes in 2009 or beyond.
Here is an example (http://www.intel.com/references/pdfs/Turtle_Entertainment_V3.pdf) where by maximizing their consolidation with a 4S server an IT customer saw a 30+% reduction in TCO while improving their customer's gaming experience.
HP has a problem in this new market condition... Intel has consistently been late to market with every Itanium to date and in general the capabilities have been less than promised.
Case in point Tukwila was promised at 8-16 Cores with a 65nm process in 2005. It was expected in 2007. Tukwila is still not out and is only expected to have 4 cores. Montvale was supposed to come out at 2.8GHz, but only came out at 1.66GHz and two years late.
Intel has a history of being late with Itanium and there's no reason to believe this won't continue. Perhaps it's because Itanium is way too complex and difficult to advance. They took their own route rather than RISC and X64, Power and even SPARC are going to kill Itanium (which has met no ones expectations thus far).
HP's in trouble in the enterprise market.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
