In my quest to get a handle on where servers are headed in 2010, I've spent time thinking about architectural innovations from Intel and AMD. I've also been serially interviewing the server vendors themselves (see my new piece on HP). Now comes the next step--I'm pulling together
In my quest to get a handle on where servers are headed in 2010, I've spent time thinking about architectural innovations from Intel and AMD. I've also been serially interviewing the server vendors themselves (see my new piece on HP). Now comes the next step--I'm pulling together a survey for InformationWeek Analytics. And I'm asking for your help.What I'm requesting is your input on which issues are of key importance to you in the server space. I don't care if you're a specifier who buys equipment for your organization, an admin, or a user. What I'm most interested in is your insight on what I might be missing, perhaps because it's an emerging issue that's low on my radar screen. (Or it could just be that I'm just clueless. . .)
Here are the areas already on my list:
Virtualization. As in, how important are the virtualization capabilities of the server? Mostly, this is a software-side implementation issue, but it relates to the hardware in terms of its processing power as well as how many cores and sockets are on the motherboard. (Every physical core can be virtualized into an additional N logical processors.)
Power and Cooling. How big an imperative is it to save electricity? How heavily are your facilities people leaning on you?
Consolidation. How does the need for extracting efficiencies--aka reducing box count factor and saving operational costs--factor into your acquisition decisions? (This obviously overlaps with the power and cooling question.)
Processor architectures. The big battle here is Intel (Nehalem, Westmere) versus AMD (Opteron Istanbul). I've neglected the high-end RISC architectures such as Itanium and UltraSparc; those obviously have a place, too.
Storage and Networking capabilities. Support for 10-GB Ethernet. Fast I/O out to storage and flexibility in connecting to SANs.
Management Software features. Importance of the automated ability to discover, allocate, and manage resources.
Server mix in your data center. What's the distribution of commodity, mid-range, and high-end systems?
Memory config. Given the drive towards virtualization, what's the sweet spot for memory? The assumption is that everyone's specifying a lot more than they used to.
Where does the server buy fit, in relation to a complete data-center architecture? Many vendors now are positioning their boxes as part of a holistic, converged architecture. I often wonder, the good things about this approach notwithstanding, whether this sometimes makes it more difficult to parse the particular merits of the processing portion of the equation.
What am I missing? Let me know, by leaving a comment below or e-mailing me directly at firstname.lastname@example.org.