The state of HPC tier 1 vendors

Much has been happening in the HPC tier 1 vendor space. Some of it has made the news, much has not. The TL;DR version: I believe that most of the tier 1 HPC capability may have been wiped out over the last few months. 1 tier 1 and a bunch of tier 2 are left. Basically, the HPC market has a number of tiers within it, and product mixes across these tiers. In the HPC cluster space, IBM, Dell, and HP have almost completely dominated systems sold … not necessarily in terms of the top500, but in a broader sense, with a small fraction by lesser known entities. Cray bought Appro a while ago, and if you look at the known public details of the purchase, this was at a low valuation. It was basically an acqui-hire. My old employer (left them 13 years ago) SGI has a few systems as well. Clusters are also very much in demand. They are known quantities to companies and universities. People know how to manage them. And they aren’t all that far removed from “clouds” though the hyperscale clouds tend to look very … very different from HPC clusters. So with IBM tossing out its x86 (well, over to Lenovo) and the various complications this presents to public sector purchases in the US, it also presents an interesting moment for private sector purchasers to reflect upon the state of the tier 1 players. IBM: sold bits to Lenovo. The pros are that Lenovo is a persistent force, and they will likely figure out how to make money out of these systems. The cons are, that the sweetheart deals IBM created for many in the past, very low or negative margin deals, are likely gone. Also a con, the teams they’ve built up for selling these systems are also likely gone. IBM’s HPC offerings no longer include clusters, they are now Lenovo clusters, and in the era of very tight margins on clusters, Dell: Before going private, they had been quietly shedding people, including a number of friends and former colleagues. I’ve heard some rumors on the state of their HPC practice, but I am not yet ready to talk about that. But this should be a pointed question when engaging with them. One’s commitment to a particular markets needs can be seen in their hiring and marketing practices for larger companies. Smaller firms live and die in single or small numbers of markets, but the bigger companies enter/leave/create/shutdown opportunities in markets by their actions. Their commitment to a market is measurable in terms of what they keep and what they jettison when times get challenging. HP: Has been trying to get out of x86 for years now. They hired a failed CEO, and replaced him later after he largely blew up lots of business with his inept handling of delicate market transitions. But his efforts exposed to the HP customer base, some of the internal deliberations that they were having. I am not sure that Ms. Whitman can fix this. And in the bigger picture, does it make sense to if you are planning on leaving anyway? Whom are the other tier 1? I’d argue SGI is in tier 2. They aren’t as big, and the company as it exists today bares almost no resemblence to the former HPC powerhouse it once was. Cray with Appro is a tier 2 for clusters, though they have a good gig going with the national labs, where they might be considered a tier 1. If you think about this, this leaves all of the smaller tier 2 players with a wide open playing field. The Lenovo and Dell revamped units aren’t likely to do the same low margin deals that many expect in this space for clusters. I mentioned something about this in the dim and distant past … 2008 actually.

Now some of the most “competitive” market dwellers appear to be leaving this market, after driving out the innovative folks (with many of the tier 2 being reduced to little more than box resellers), choice is reduced, and there are some interesting future consequences of this. Which I think will be seen in subsequent purchases. Basically the large tier 1’s need to focus themselves on what they perceive to be their growth markets moving forward. Which means that they can’t commit to markets that aren’t core to them. And in order to streamline for the push into new markets, they need to trim their staffs. Thats quite a lot of HPC talent being kicked to the curb. I suspect that Cray, SGI, Penguin, and other tier 2’s are going to start pushing very hard. All of these resell Supermicro gear though, and as I noted recently, it can’t be far from anyones mind that they (Supermicro) could be acquired, and the rest of that market taken over. Just look at what Netapp did with LSI Enginio. Suddenly a large swath of the storage market were reselling rebranded Netapp. Same with the Xyratex purchase by Seagate. Seagate is now actively competing with some of its customers. So the tier 1 in HPC is being rapidly swapped out for tier 2, there is interesting future “risk” that these tier 2’s could be swatted out of the way by whomever purchases Supermicro … or Supermicro itself could decide to move more up market. FWIW, we’ve seen evidence of the latter already. We do live in interesting times …