Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.
January 4, 2010
3 Min Read
In AMD's case, I'd characterize the Istanbul architectural enhancements as taking to the next level some of the stuff AMD pioneered when it debuted the first multicore Opterons back in 2005. Of course, this list includes hardware virtualization support and advanced power management. But we're also talking extremely fast processor-to-memory bandwidth, via AMD's Hypertransport link. (The original Istanbul press release is here.)
An interesting angle is that the Istanbul Opertons were pin-compatible with existing motherboards, enabling them to be used as drop-in upgrades.
AMD took its Istanbul strategy to the next level in September, announcing a new server platform spec, code-named Kroner. The salient feature here -- aside from giving a boost to AMD's home-grown core-logic chipset for Istanbul, one of three chipsets available -- is the low-power angle. The ability to hang a low-power tag on one's server is going to be a must-have checklist item to sell into the data center market in 2010.
My overall point here is that both Intel and AMD have given system builders a more advanced processor pallet to work with than ever before. So while I don't mean to suggest that we're going to see shockingly different servers from each vendor, I do believe this means there will be more differentiation at the margins. Low power or power efficiency, or whatever you want to call it, will come first. (Yes, I know that "efficiency" has a technical meaning, and is commonly misused.) But this differentiation will extend to other areas, notably ease of management.
For example, Dell, in a December press release, touted its offering of an Infrastructure Manager as part of its servers-for-the-data center solution. The manager, wrote Dell, "enables customers to dynamically allocate workloads in minutes by altering server, networking, and storage devices without the need to re-cable, reconfigure, or reload software."
The Infrastructure Manager runs on a (Nehalem) Xeon 5500-based PowerEdge M610 Blade.
Then there's Hewlett-Packard, which has taken an interesting approach to the energy-savings play via taking some of its ProLiant servers "skinless." These aren't barebones in the processing department, but rather eschew the traditional rack for a lighter tray carrier. This enables HP to bundle them into something it calls the ExSO, or Extreme Scale-Out, portfolio. It's positioned as a modular data center solution. (I should note that other vendors also have similar, skinless rack-type products.)
In closing, I should note that perhaps I'm conflating vendor differentiation with a broader trend. That would be bundled packages positioned as one-stop data-center solutions, all under the umbrella of competing with, and stealing the center of gravity from, Cisco's Unified Computing System. UCS is a brilliant marketing method of pulling everything under an "easy-to-manage" umbrella. It's almost hard to beat, even with a best-of-breed argument.
For Further Reading
What's your take? Let me know, by leaving a comment below or e-mailing me directly at [email protected].
Follow me on Twitter: (@awolfe58)
Alex Wolfe is editor-in-chief of InformationWeek.com.
About the Author(s)
You May Also Like