GlusterFS or Ceph: Who Will Win the Open Source Cloud Storage Wars?

The open source cloud storage wars are here, and show no sign of stopping soon, as GlusterFS and Ceph vie to become the distributed scale-out storage software of choice for OpenStack. The latest volley was fired this month by Red Hat (RHT), which commissioned a benchmarking test that reports more than 300 percent better performance with GlusterFS-based storage.

Actually, it might be most precise to describe the GlusterFS-Ceph competition not just as a war, but as a proxy war. In many ways, the real fight is not between the two storage platforms themselves, but their respective, much larger backers: Red Hat, which strongly supports GlusterFS development, and Canonical (the company behind Ubuntu Linux), which has placed its bets on Ceph. (Ceph itself is directly sponsored by Inktank, which has a close relationship with Canonical, and in which Ubuntu founder Mark Shuttleworth has invested $1 million of his own money.)

Red Hat’s latest effort to pursuade its customers and the open source community to embrace GlusterFS comes in a benchmarking report by Principled Technologies. The testing, which Red Hat commissioned, compared the performance of OpenStack running with Red Hat Storage Server using GlusterFS and a Ceph storage server. It found that read throughput ranged from 2 times to 3.8 times as fast with Red Hat Storage versus Ceph, depending on the number of compute nodes involved.

Ceph advocates might point out that, as the report notes in its appendix, the Ceph server involved in the testing was running on Red Hat Enterprise Linux 6.4, while Red Hat Storage Server is a standalone Red Hat platform—so the report didn’t completely compare apples to apples. The version of Ceph used for the tests, 0.67.4, is also not the latest available (that’s 0.72).

But these details don’t really matter from the channel perspective. The real story here is the larger battle between Red Hat/Gluster on the one hand, and Canonical/Intank/Ceph on the other. And what makes it particularly important is the extent to which their fierce competition diverges from the norm within the open source ecosystem, where in most cases a single core technology dominates. For example, OpenStack is the only game in town when it comes to building clouds, while Hadoop is the near-universal solution for Big Data. And Linux, of course, is at the core of it all; even though an array of different Linux distributions exist, they share a common core in the Linux kernel, which they all support.

Yet this kind of pan-channel cooperation does not exist in the niche of scale-out distributed storage. Perhaps, sooner or later, either GlusterFS or Ceph will win out, and the open source ecosystem will have a single cloud storage solution around which the various players can build their platforms. But for now, the battle rages.