Analysis: Would you like a data center with that server?
- — 20 January, 2010 08:03
Faced with the continued commoditization of servers, IT vendors this year will try to differentiate their offerings by moving toward more highly integrated, unified compute platforms.
That means combining servers, storage and networks -- and related management software -- as a package deal, says Rockwell Bonecutter, data center technology and operations practice lead at Accenture. Vendors tried something similar a few years ago with the launch of modular "data center in a box" products. But this time, they are thinking out of the box by focusing squarely on the core data center infrastructure.
Soup to Nuts
With integrated systems, you're not just buying a server; you're buying the entire infrastructure to go with it as a single, managed entity. The systems are highly integrated and highly scalable, and they include computing, storage and networking systems, as well as business services management, automated orchestration and provisioning capabilities. In some cases, those systems might include power delivery and cooling as well.
Bonecutter says the approach addresses real pain points for IT. "I'm spending the majority of my time meeting with CIOs about implementing cloud computing or [about] orchestration and provisioning," he says.
HP's BladeSystem Matrix is a harbinger of what's to come, says Jim Ganthier, vice president of HP's industry standards group. He says Matrix lets administrators remove and reinsert blades between slots, while virtual connects ensure that storage and network resources move with them -- no reprovisioning required.
The systems also allow for workload balancing based on the need to optimize power consumption, usage or a combination of the two. With Matrix, he says, "your enclosure is your data center." But HP plans to implement the same model this year using rack-mounted HP servers with storage and switch gear.
IBM has a similar approach. It wants to create a more efficient, integrated system that weaves everything together more tightly. But will CIOs be willing to buy all their storage, server and networking equipment from HP or IBM to make it work?
Tom Bradicich thinks so. The IBM fellow and vice president for architecture and technology for IBM's x86 servers says the interchangeability, or "fungibility," of the x86 platform is what made it so popular.
He says enterprises will be willing to trade some fungibility for a more efficient system -- albeit one in which the x86 system morphs to include some proprietary technologies and the customer buys the integrated package from IBM.
Though he acknowledges it will be "less fungible," he believes "that will be tolerable because complex workloads make that trade-off worthwhile. There will be a higher tolerance for some parts of the system to be more fixed than fungible."
Dell is pursuing a similar strategy but is weaving together support for third-party equipment and taking a more open, standards-based approach. "We use heterogeneous components but manage in a converged fashion," says Paul Prince, chief technology officer for Dell's enterprise products group.
Robert Beach, director of IT services for the Seminole County, Florida government, says the idea is an interesting one, but he's not sure he wants to be locked into one vendor for everything. "To me, it sounds like you're putting a lot of eggs in one basket," he says.
Scottrade CIO Ian Patterson, who runs a very large data center, also has concerns about fungibility and vendor restrictions. But, he says, "it could also be a positive in the long term, simply because of the ability to monitor the interactions of all of these pieces from one box."
He sees the potential to get better performance than he gets with devices managed across the network, but he says it's still too early to judge. "I would look out at least 12 months before investigating this as a viable technology," Patterson says.