On Monday the news was abuzz with word from the Microsoft Partner Conference—tales of its various partnerships to enhance its Azure cloud offering were proliferating, complete with the requisite surprises and some items that dropped off the radar as soon as I heard them given our HPC focus. Monday night, I went to bed comforted by the notion that when I would wake up on Tuesday, for once I would have a predictable amount of news coming directly from the Redmond camp. I wasn’t sure what partnerships they would declare or how they planned on bringing their private cloud in a box offering to real users, but I knew that I would be writing about them in one way or another.
Little did I know that the Tuesday news I was expected didn’t surface at all and in its place,this remarkable announcement from Amazon about its HPC-tailored offering, which they (rather clunkily) dubbed “Cluster Compute Instances” –(which is thankfully already being abbreviated CCI).
I almost fell off my chair.
Then I think I started to get a little giddy. Because this means that it’s showtime. Cloud services providers, private and public cloud purveyors—everyone needs to start upping their game from here on out, at least as far as attracting HPC users goes. If I recall, Microsoft at one point noted that HPC was a big part of their market base—a surprise, of course, but it’s clear that this new instance type (that makes it sound so non-newsy, just calling it another new instance type) is set to change things. Exciting stuff, folks.
As you can probably tell from this little stream of consciousness ramble, this release today turned my little world on its head, I must admit. I hate to be too self-referential, but man, I wish I could back now and rewrite some elements of a few articles that have appeared on the site where I directly question the viability of the public cloud for a large number of HPC applications in the near future, including those that invoke the term “MPI”—Actually though, it’s not necessarily me saying these things, it’s a host of interviewees as well. The consensus just seemed to be, Amazon is promising and does work well for some of our applications but once we step beyond that, there are too many performance issues for it to be a viable alternative. Yet. Because almost everyone added that “yet” caveat.
I did not expect to see news like this in the course of this year. I clapped my hands like a little girl when I read the news (which was after I almost fell off my chair). It is exciting because it means big changes in this space from here on out. Everyone will need to step up their game to deliver on a much-given promise of supercomputing for the masses. This means ramped up development from everyone, and what is more exciting that a competitive kick in the behind to get the summer rolling again in high gear on the news-of-progress front?
Microsoft’s news was drowned out today—and I do wonder about the timing of Amazon’s release. Did Amazon really, seriously time this with Redmond’s conference where they’d be breaking big news? Or was the timing of this release a coincidence? This one gets me—and no one has answered my question yet about this from Amazon. What am I thinking though; if I ask a question like that I get a response that is veiled marketing stuff anyway like, “we so firmly believed that the time is now to deliver our product to our customers—just for them. Today. For no other reason than that we’re just super excited.”
We’re going to be talking about this in the coming week. We need to gauge the impact on HPC—both the user and vendor sides, and we also need to get a feel for what possibilities this opens up, especially now that there is a new player on the field who, unlike the boatload of vendors in the space now, already probably has all of our credit card numbers and personal info. How handy.