Feeds

Cray ends battle with Appro by ingesting it alive

If you can't beat 'em, eat 'em

Top 5 reasons to deploy VMware with Tegile

Cray is going to top off a pretty good year financially if it can get the many-petaflopping Titan and Blue Waters supercomputers accepted at Oak Ridge National Laboratory and the University of Illinois.

There's a good chance those machines will be accepted, but it creates a problem: the $188m Blue Waters deal was an unexpected win, and next year, the company will be facing down a very tough compare. Moreover, Cray doesn't really have a plain vanilla x86 cluster business, and particular does not have one that is suited for helping it build Hadoop and other big data iron.

And so, to kill several birds with one stone, Cray has decided to buy HPC rival Appro International for $25m in cash.

Having sold off its Gemini and Aries interconnect development team and technology to Intel in April for $140m, the company has ceded the development of interconnects to Intel, Mellanox Technologies, IBM, and a handful of much smaller and niche players in exchange for a big wad of cash.

The acquisition of Appro fits in with Cray's plans to expand outside of its core "capability systems" business, where a couple of high-end deals make or break the company.

Peter Ungaro, the CEO at Cray who used to run IBM's HPC business, has done a remarkable job in getting Cray off the rocks and winning big deals and a few small ones. He explained in a conference call with press and analysts that Appro was "kind of a missing part of our puzzle".

What Ungaro meant is that Appro sells clusters based on rack and blade servers lashed together with Ethernet or InfiniBand networks instead of the exotic SeaStar, Gemini, and Aries interconnects that have defined Cray systems for the past decade. Appro also uses Intel motherboards, rather than custom boards that Cray systems have, so its systems are cheaper.

Don't be fooled by the daft name, the Xtreme-X makes sense for big labs

For some customers, an Appro Xtreme-X super is the only kind of machine they will buy. And that is what has made the company the number three supplier of high-end systems in the supercomputer market, as reckoned by the Top 500 rankings.

On the June 2012 Top 500 list, Appro had 19 systems in total; its largest machine being the Zin cluster at Lawrence Livermore National Laboratory, which has a Linpack rating of 774 teraflops sustained. Sandia National Laboratories and Los Alamos National Laboratories also have big Xtreme-X clusters.

Ungaro said Cray had been looking around to buy a number of HPC suppliers with generic cluster expertise, and that Appro "really seemed to be the best fit" because the company takes an "engineer-first" rather than a "marketing-first" approach to building and selling systems.

Appro is privately held and therefore does not release its quarterly financials, but the fact that Cray is expecting Appro to add around $60m in revenues to Cray's top line in 2013 and it is only paying $25m for the company is a pretty good indication that the high-end of the HPC market is very tough and margins are not great even if they are worthy of being in business.

Appro from nothing

Daniel Kim founded Appro International in May 1991 and was still CEO at the company until he sold it to Cray today. Kim initially focused on selling rack-mounted servers, and got a big share of that market before all the tier-one players crowded in.

A decade later, when they were all selling racks like crazy, Kim shifted Appro into a higher gear, focused the company on building high-end supercomputer clusters. Since that time the company has invested in hardware and software engineering to build a respectable business. Appro is headquartered in Milpitas, California and has a supercomputing center in Houston, Texas. The company has manufacturing operations in South Korea and Japan.

Brian Henry, CFO at Cray, said the Appro acquisition was not expected to require approval by the US government under the Hart-Scott-Rodino antitrust act, and could close in a matter of days or weeks. The deal requires Appro have $3.5m in net working capital and no debt to get the $25m, and if it doesn't, then the amount of cash is adjusted down to compensate.

After the deal closes, Kim will become manager of the Cluster Solutions business at Cray, which will have its own research, development, product management, and marketing teams (coming from Appro itself). The Appro sales team will be merged with the Cray sales team and will sell all Cray solutions. Cray will bring at least 90 employees from Appro. Henry said that the addition of Appro to the Cray business would be accretive to profits.

In addition to announcing the Appro acquisition, Cray also fired off its financial results for the third quarter a bit early. In the quarter ended in September, Cray booked $35.7m in revenues, down 2.6 per cent year on year. Product sales were up 14.5 per cent to $18.3m, while services revenues were down 15.9 per cent to $17.4m.

In the past few years, with big deals all pushed to the back-end of the fourth quarter, the third quarter has been quite weak for Cray, and this time around was no surprise. Neither was a net loss of $5.1m, which was at least lower than the $12.3m loss Cray had in the year-ago period.

For the nine months ended, Cray's sales are up 61 per cent to $232.2m, thanks to some big sales that were supposed to close in Q4 2011 being pushed out into third year, and if Cray gets acceptance on the Titan and Blue Waters supers, it should hit somewhere between $430m and $450m in sales this year.

That Blue Waters deal could add as much as $150m to the top line for 2012, Henry said on the call, and warned Cray does not have any orders of that magnitude on the books for 2013.

So the compares are going to look pretty bad. The Blue Waters bonus only happened because IBM backed out of building a Power7-based behemoth for the National Center for Supercomputing Applications. Henry and Ungaro said several times on the call that the core HPC business was growing nicely and was expected to continue to do so in 2013 with the launch of the XK7 Opteron-Telsa hybrids and the new Xeon-based XC30 "Cascade" systems.

Cray ended the third quarter with $258.9m in cash and equivalents, which includes the $25m it will spend to buy Appro, and no dents. It would not be at all surprising to see Cray make some more acquisitions, particularly in the software side of the HPC and big data space. ®

Beginner's guide to SSL certificates

More from The Register

next story
It's Big, it's Blue... it's simply FABLESS! IBM's chip-free future
Or why the reversal of globalisation ain't gonna 'appen
'Hmm, why CAN'T I run a water pipe through that rack of media servers?'
Leaving Las Vegas for Armenia kludging and Dubai dune bashing
Microsoft and Dell’s cloud in a box: Instant Azure for the data centre
A less painful way to run Microsoft’s private cloud
Facebook slurps 'paste sites' for STOLEN passwords, sprinkles on hash and salt
Zuck's ad empire DOESN'T see details in plain text. Phew!
CAGE MATCH: Microsoft, Dell open co-located bit barns in Oz
Whole new species of XaaS spawning in the antipodes
AWS pulls desktop-as-a-service from the PC
Support for PCoIP protocol means zero clients can run cloudy desktops
prev story

Whitepapers

Cloud and hybrid-cloud data protection for VMware
Learn how quick and easy it is to configure backups and perform restores for VMware environments.
A strategic approach to identity relationship management
ForgeRock commissioned Forrester to evaluate companies’ IAM practices and requirements when it comes to customer-facing scenarios versus employee-facing ones.
High Performance for All
While HPC is not new, it has traditionally been seen as a specialist area – is it now geared up to meet more mainstream requirements?
Three 1TB solid state scorchers up for grabs
Big SSDs can be expensive but think big and think free because you could be the lucky winner of one of three 1TB Samsung SSD 840 EVO drives that we’re giving away worth over £300 apiece.
Security for virtualized datacentres
Legacy security solutions are inefficient due to the architectural differences between physical and virtual environments.