Open standards for blades? Sure. Interoperable standards inside the chassis? No thanks.
That was the bottom line yesterday during a panel discussion on blade server standards and innovation at Computerworld's Data Center Directions virtual trade show.
What's great about standards in the server blade realm is that there are plenty to go around. Every vendor has its own "ecosystem." And while leaders IBM and HP are fine supporting standards outside of the blade system chassis, inside the box the systems remain proprietary. The specifications for each design are open in the sense that vendors of add-in devices (storage and network switches) that want to play inside the box can do so, but the vendors play by IBM's or HP's rules in that space. In each case they must create a custom version of their product that's plug and protocol compatible with each blade system vendor's box.
One could get around this by leaving critical IO elements outside of the chassis, but that can create bottlenecks as the number of blades - and virtual machines - per chassis continues to climb. The chassis backplane is the new server expansion bus.
Intel's Server System Infrastructure initiative (SSI) promotes open, modular specifications for some elements of blade server designs, but so far both IBM and HP have held back in favor of their own designs. SSI provides for modular, interoperable "building blocks" for server blades, chassis designs and manageability software. "Why is everything proprietary to a box? There are folks that want more open standards even inside the chassis, says Kurt Lender, senior blades marketing manager at Intel Corp.
But Gary Thome, Director, BladeSystem Strategy and Architecture at HP, argues that the world inside the chassis should remain a proprietary space where vendors can innovate and differentiate in such areas as power management.
Alex Yost, Vice President for the IBM BladeCenter, says SSI is for the smaller, niche market blade system vendors that lack a large enough installed base to create their own ecosystem, not for IBM, which has a tremendous amount invested in its own infrastructure and designs - and has plenty of market share to have its own way.
As the blade server market has matured it's striking that the server world seems to have moved from an interoperable, industry standard backplane like PCI to a series of open, but proprietary backplane and connector designs into which one must purchase and install vendor-specific versions of every IO device.
Today, the only industry standard expansion card support in server blades typically consist of a single mezzanine card slot side-mounted on each server blade. The critical "expansion card" slots are in the chassis backplane, which means you'll never be able to move that Cisco or Brocade switch installed in a BladeCenter into an HP BladeSystem c-class chassis - or vice-versa - as your needs change.
Then again, perhaps it doen't really matter. Both IBM and HP left me with the distinct impression that all this is nothing more than a tempest in a teapot.
Are plug and play standards important to server administrators anymore inside the server box when that box is a server blade chassis? Or is innovation king and interoperability at this level irrelevant, so long as everything is interoperable outside of the chassis?