by Senior Software Architect Nathan Rutman, patent holder of numerous recent Xyratex software inventions. Part 1 of 2 briefly goes into the advantages of open source Hadoop software architecture that is dominating enterprise big data applications and which Xyratex is using in a number of its new racks(eg Pentaho, AMD,Cray Big Data). Blog then goes into significant weaknesses of Hadoop as it applies to Humongus Data particularly overall cost of large systems hardware(TCO) and speed(I/O). To say this blog is technically over my head is a vast understatement.
But I have heard some of these same sentiments before while listening to occasional Lustre panels discussing advangages of Lustre over Hadoop for enterprise. And not just Xyratex Lustre people but others from throughout that community believe that in the not-too-distant-future, Lustre updates will make it enterprise ready. Hadoop is much easier to use and why its probably not a big deal for Xyratex to use it in their racks. Lustre is too complicated now for enterprise. Many say within a couple years the intense Lustre evolution going on now will fix all that and Xyratex will be at the fore-guaranteed. With a lot of brilliant minds like Nathan, the Peter B's(x2) and many others that we've seen and heard from on occasion it's no surprise Xyratex is not only becoming a goto guy in world-class storage but also hard charging leaders in the Lustre community. Watch out Hadoop, there's a bigger faster brute on the way. When this happens Xyratex will be aiming for more than 5% of the Big Data/HPC/Cloud markets. A lot, lot more.
BTW, in two weeks Xyratex' own Peter Braam, The inventor of Lustre, will be a featured keynote speaker at Lustre User Group-2013