I didn’t really get it, back in January 2014, when Intel, Amazon and others made a big noise about slapping the equivalent of an ‘Intel Inside’ sticker on your favourite cloud.
Now they’re at it again, with the birth of a new ‘Cloud for All’ initiative, a nudge for OpenStack, and the promise that Intel and Rackspace will – in six months – deliver two 1,000-node test clusters. Just wait for someone to comment that OpenStack/ Intel/ Rackspace must be hard/ slow/ lazy if it takes six months to walk down the racks pushing 1,000 ‘on’ buttons…
I still don’t get it.
Test clusters are good. Intel, Rackspace, and anyone else who wants to work to optimise code, smooth wrinkles, fix problems, validate conformance and all the rest of it? Also good.
But the whole Cloud for All thing? Not so much. Lots of cloudy stuff runs on Intel. Lots doesn’t. Lots of cloudy stuff runs on OpenStack. Lots doesn’t.
Intel was already working to make sure its chips and other gizmos suited the needs of their customers. Operators of clouds are one, increasingly important, class of customer. Intel was already collaborating with OpenStack, and with partners like Rackspace.
Giving existing collaboration a fancy name doesn’t make it more important, or better, or anything. It just looks a bit desperate.
In fact, if they weren’t already doing most of this stuff, heads should have rolled long ago. Doing what you should be doing isn’t the stuff of headlines and press releases. It’s the stuff of business as usual, behind the scenes, quietly, without fanfare. We only care when you’re not doing what you should be doing. When you’re doing what you should be doing, and you shout about it, we all go, “Well, d’oh.”
As I mentioned in my post about last year’s announcement with Amazon, this all seems such an odd thing to shout loudly about. Cloud providers care deeply about increasing efficiency, decreasing energy consumption, and more. All of that goes straight to their bottom line. But their customers couldn’t care less, most of the time. Those customers don’t require an Intel chip, and they don’t require a partnership, and they don’t require squeezing a little more efficiency out of a physical server. They care about the virtual machines and the applications that are delivered to them. They care about how well they run. They care about how much they cost.
They don’t care how the cloud provider makes that happen, so long as they do.
(Cross-posted @ Paul Miller)