"Evolution" for Microsoft HPC
By joe
- 8 minutes read - 1634 wordsThis is old news at this time, but Microsoft has moved its HPC group into their Cloud groups. I’ve talked in the past about critical business decisions that need to be addressed over time, as a business matures, and a product line is given time to sink or swim. At the end of the day, a business has to make hard decisions about what products to introduce, which to end-of-life, which to grow independently, which to fold into other initiatives. In a general sense, businesses exhibiting rapid growth, you support, and support hard. You give them resources, you feed them additional capital. You don’t fold them into other initiatives.
Understand, I’ve been critical of Windows for HPC, in part because of its unlikeliness of being a real profit center for Microsoft. It had looked to me as a tactic in a broader strategy against a competitor, Linux. I know some may disagree with me on that characterization. Thats fine. But it is still quite apt. And that is how it was played by Microsoft, at least indirectly. And that strategy failed. Well, more to the point, the cloud happened. And cloud is pretty much non-Microsoft. Apart from Microsoft’s cloud, and a few others in isolated spots. The argument I make is one that I think the business stats bear out. There was some sort of management led ego-based war against specific competitors going on, and leadership took its collective eyes off the ball. Microsoft wound up not out in front innovating, but lagging. In a sequence of markets that changed rapidly. On the client side, tablets are pretty likely to all but replace laptops over the next few years. Lighter, cheaper, do most of what people need. Most don’t run windows, and most don’t run office apps. This is a massive … and existential … problem for Microsoft. Google has their apps, and regardless of whether or not they are as good as Microsoft’s, they are … good enough. You don’t need client installed apps any more. And Apple is developing their own tools, both client side and cloud side. And making their units untethered, storing everything in the cloud. Desktops are being virtualized, and the desktop will be visible on the tablet. I’ll argue this will be a transitional period. Before everything moves to the tablet. And where will Microsoft be, given that they don’t make the OS that runs the tablet, and that they don’t make the apps that work on the tablet? Yeah, this is why taking your eye off the ball is a really bad thing. This is why going on an ego-fed war against a competitor, rather than embracing the competitor and leveraging their capabilities to grow your market, was a very … very bad move for them. And then we come to HPC. HPC won’t be done on tablets … well, the computing side anyway, unless NVidia or someone else starts putting 100+ cores into a tablet. No one wants a warm tablet that has a fan and 30 minutes of battery life, so I expect most computing to be done on remote systems. Will these be Microsoft systems? Probably not. Remember, the claim was Microsoft was going to democratize HPC. Well, it looks like iOS and Android are taking over the tablet and phone markets, with windows found only on the dying Nokia, and a few other places. Phone vendors aren’t dumb, they see what sells. Tablet vendors aren’t dumb, they see what sells. These systems have democratized the client side of things. No longer is there a single monopoly controlling most of the client base. Now there are 2 dominant, and if HP pushes hard, 3 OSes people can choose from. Yeah, I like Blackberry and RIM. Love their phones. Some of the best on the market. Unfortunately, they look like they are driving away all their developers. HP’s WebOS is a Linux stack. It does look good, but I am a little wary of their being a single vendor behind it. Similar with Apple, but I don’t think Apple would kill iOS as easily as I think HP could kill WebOS. Nokia is, sadly, toast. I’d argue that they hired the wrong person as their CEO, and he is going to be the person who shuts out the lights. I’ve seen this, and worked at companies before, where really … insanely … bad decisions were made, and followed through, to the grave detriment of the company, its stockholders … Think SGI and its 2 bankruptcies. What has this got to do with HPC? The HPC division was folded into the cloud division. This is a way to preserve the HPC marketing bits without allowing a claim that the division was closed, and to continue product support. But it also bodes ill for the stand-alone HPC product. Microsoft now acknowledges that cloud is important. Unfortunately, for every cloud server instance, there is one less physical installation. And from a cost perspective, open source OSes and apps mesh perfectly with cloud systems. Per CPU cost OSes don’t. If you want to launch a 1000 node cluster to work on a problem, you don’t pay for 1000 licenses of an OS to light up to run your cluster. No, this is done via open source technologies, so the pricing is flat. Though if you want to pay Redhat for support in the cloud, you can. Which works well for them, as their model works remarkably well for physical or virtual servers. Less so for pure pay per instance servers. Your costs will always be higher. Survey Amazon and other cloud providers to see. Windows machine rentals cost more. If your app doesn’t care, which will you consume? Which is why cloud is a more serious problem for Microsoft than it might actually wish to admit. Client machines going away are an existential attack on its revenue base. Cloud always being more expensive is a fundamental challenge for them for the future. So how does HPC play in this? Right now, various public clouds are used for a number of HPC calculations. Those that are less sensitive to latency and other issues. Or those who need less IO bandwidth. Yeah, they work well in the cloud. Think massively parallel apps. I can’t imagine that this is a viable business strategy for HPC at Microsoft going forward. What I can see is that they needed a way to walk back from a commitment, and do so carefully. So a reorg and a fold-in. Gradually, HPC “features” will make their way into the other products, and the group will gradually melt away. Allows them to avoid answering some tougher questions. This said, I’ve seen (and we’ve sold) very few Microsoft HPC installs. We’ve seen and heard of very precious few. As I’ve noted many many times, its always a business decision. The same machine with a Linux distribution on it will run as fast and be as easy to manage, but will cost less. Now run larger ones. The economics never worked out. The can’t work out given the business model for the product. Cloud is only going to exacerbate this, and the client attack is going to require a deep re-thinking at the company. As of today, Apple is the largest US company by market valuation, and I don’t see this changing soon. We (as in me personally) am considering getting a Mac for my desktop, to run along side my Linux machine. Nothing against my Linux machine, but given what I’ve seen of the Linux desktop situation, with Ubuntu trying to be a tablet OS, the relevant groups working on discarding X and other things … it makes me wonder how much longer I should use it as my primary desktop. Maybe 15 years of doing so is enough. I dunno. Microsoft needs to focus. It needs to think carefully about what it wants to be in 1 year, 3 years, 5 years, 10 years. The trajectory its on and that it has been on is not the one it wants to be on. It needs to cut things that are non-core, and figure out what is core. I’ll argue, HPC isn’t core for it. Not even close. It really has far bigger fish to fry. Focusing upon HPC would be a distraction. Focusing upon the emergent existential challenges … would be wise. So folding the HPC group into the Cloud group makes sense in this context. It is a good business decision, allows them to “continue” the project for a while, avoid overt cancellation, and the subsequent eating of crow. They’ll update the software for a while, and make a few new releases. And the team will gradually be given other priorities, and the functionality merged into other areas. Which again, allows them to declare “victory”. Even if it technically isn’t. Everyone I’ve met or spoken with on their HPC team has been quite nice. Very enjoyable people. Even when I disagreed with them on fundamental aspects (such as “does it make sense for Microsoft to be doing this”), they engaged in a professional and lively debate. I don’t think they will agree much with this post. Thats fine, and I invite them to comment, even anonymously, and I’ll vigorously preserve and protect their anonymity. I don’t think it made sense to pursue HPC, but then again, I had a sense over the last decade that Microsoft wanted to be all things to all people. For a time it was. I believe that time has past. So HPC as a stand alone is no more, and it is now part of Cloud/Servers. Thats the take home message. What happens after this … well, my prediction is on record. Lets see what happens in a 1,3,5 years.икони на светци