What does SGI want with LNXI, anyway?

Print Friendly, PDF & Email

HPCwire has a story from Tabor analysts Addison Snell and Chris Willard with some thoughts on what in the hell anyone on either side of this deal was thinking.

Before we get started on that, though, I wanted to ask about this line from the story

SGI struggled famously for 10 years and emerged from bankruptcy vowing to reclaim its prior HPC status, and indeed, with former LNXI CEO Bo Ewald in charge, SGI is showing signs of improvement.

I’m sincerely not being snarky here, but how, exactly, is SGI showing signs of improvement? I think the company has some great technology that we need in this business — it’s the only vendor producing large scale shared memory machines.

But the company’s string of quarterly losses remains unbroken, and they actually widened this quarter. The company has announced that it’s getting back into visualization, but that business is pretty slow to get started. The ICE offering was already on the table when Bo arrived. So, other than the company being in honeymoon because they have a new chief, what is better? What’s your take?

Anyway, Tabor’s view on what SGI gets out of this? LNXI’s Clusterworx Advanced cluster management suite, and its engineering staff (if they accept the offers). SGI has stated it has no plans to keep any of the LNXI hardware products alive.

Addison and Chris also speculate that SGI gains LNXI’s customers, but I think these customers are totally in play. None of the vendors can ship large systems that work out of the box anymore, because they are simply too big and too expensive to test before ship. All the vendors (broadly speaking) are using what amounts to a commodity processor, so no one is locked into a system family anymore. The end result is that there is no reason for brand loyalty anymore — the products and services simply aren’t differentiated enough at the high end, and the market at the low end is still coalescing.

I do think that SGI’s management suite could use some sprucing up, something they have been working on in recent months. The Clusterworx technology may help them spackle in that hole.

I don’t think its actually a huge win for SGI, or that it will transform their business. It can be a step in the right direction, if managed properly.


  1. […] at the always interesting InsideHPC asks a very important question, that, oddly, I think I can […]

  2. […] What does SGI want with LNXI, anyway? […]


  1. I think SGI’s improving by doing less hardware development in-house, and actually OEM’ing solid solutions and packaging for customers. They’re securing large contracts by offering a total package approach, like what CSC or IBM would do for organizations (e.g. NASA comes to mind). While I’m not saying that I think that they’re doing enough, I do think that they’ve made progress. SGI needs a “killer app” and different management to get back on track though (IMHO).

  2. SGI seems to be moving past the tenuous value propositions of its past, including “TCO of an Altix is lower than a cluster” (repudiated by the market) and “SGI’s existence is critical to national security” (history says the government dole may keep a super company on life support, but won’t get it well).

    LNXI had a great product, but they were never able to compete effectively on price, and a few stumbles in execution really hurt them. The sum of the SGI+LNXI isn’t a big win but there are some benefits. LNXI understood management software, still a weak link for industrial clusters (though LNXI’s “boot off the head node” model has to go). SGI still enjoys many religiously loyal customers who have to ditch their Altix but would dearly love their cluster to be purple. Whether SGI can eat into the market share of the big players remains to be seen though.