Not long ago, the idea of deploying a server with Intel’s Atom processor would get you laughed right out of the data center. But now a startup called SeaMicro is making a compelling case for servers based on the netbook processor, and, strangely, big server vendors don’t seem to be in a hurry to counter this threat. Is it a smart move or a classic case of the competitor they never saw coming?
What We Have Here is a Failure to Innovate
IBM, HP, Dell and other server makers are doing their customers a disservice by not being more technologically progressive. Instead of helping their enterprise customers transition to a low-power, cloud-enabled future, they’re hellbent on delivering “cloud solutions” anchored on tweaked versions of their existing server, storage and networking products. It’s an understandable, if predictable, approach from a business standpoint. After all, volume servers powered by off-the-shelf Intel and AMD processors are their bread and butter, and judging by an improving server sales picture, it’s likely to stay that way for a while. So they play it safe (shocking I know).
But demand for cloud computing centers keeps growing along with the energy required to operate them. Enter SeaMicro, which arrives in time to fill the void and run off with “cloud server” mantle. How? Let’s have a look at what SeaMicro has to offer.
The Right Server at the Right Time
Back in January, I singled out SeaMicro as one of the most promising firms to receive DOE funding in the agency’s bid to improve data center efficiency. That assessment still holds today. because while large web companies continue to signal their willingness to explore non-standard server designs in search of energy savings, their pleas are falling on the deaf ears of big IT vendors.
This bodes well for SeaMicro and I’m not the only one that thinks so. As Stacey pointed out at GigaOM, James Hamilton, a vice president at Amazon Web Services and former Microsoft data center and cloud computing architect, likes what he sees. He writes:
SeaMicro have taken shared infrastructure to a [sic] entirely new level in building a 512 server module that takes just 10 RU and dissipates just under 2Kw. Four of these modules will fit in an industry standard rack, consume a reasonable 8kW, and deliver more work done joule, work done per dollar, and more work done per rack than the more standard approaches currently on the market.
By SeaMicro’s reckoning, its server (the SM1000, pictured right), which is made up of 512 Intel Atom processors, can save businesses anywhere from just over $1 million to nearly $3 million in energy costs (depending on CPU utilization). It’s an impressive return on the server’s $139,000 starting price, even if those estimates fall on the optimistic side. The picture is even brighter for data center operators that plan future builds around server infrastructures like SeaMicro’s. The reason? Power and cooling.
According to IDC, power and cooling are nearing cost parity with new server spending. For every dollar spent on a new server in 1997, it cost roughly 10 cents to power and cool it. Today that figure is nearing 75 cents. A big benefit of low-power servers is that they generate less heat. If SeaMicro succeeds in setting the standard for cloud servers, data center operators can budget for cheaper and more modest cooling systems, or if the local climate allows, free-cooling alternatives. A big win, no matter how you look at it.
And there’s a good chance SeaMicro can set the benchmark for cloud servers. Its pedigree aside (not a bad management team or roster of backers), the company boasts of the server’s “standards-based” underpinnings and makes a claim that’s music to IT execs’ ears: a low power, small footprint system that requires no modifications to existing software. If it lives up to those promises, volume server makers better start drafting roadmaps that include innovative, low-power server designs, perhaps even float a prototype or two, or they’ll risk allowing SeaMicro to become the go-to cloud server brand. Given the growth and momentum in the cloud space, that could prove to be a costly mistake.