Sign up for our newsletter and get the latest HPC news and analysis.

Inside Track: Oracle has Kicked Lustre to the Curb

He's mean. He's nasty. And he hates everything to do with Christmas!

The Bumble. He's mean. He's nasty. And he hates everything to do with Christmas!

Companies usually wind down the week before Christmas, so you don’t usually see them make a lot of strategic moves or announcements. And so it was with some marked astonishment that I received an anonymous tip that Oracle ceased development of Lustre right before the holidays. Not out of a job quite yet, Lustre engineers have reportedly been encouraged to apply for other positions within the company.

You might question the timing of this move, but what better way to bury a story than to pull the plug when everyone else is home for ten days singing Kumbaya?

No one at Oracle responded to requests for comment on this story, but that’s how the company works. When I went poking around the Lustre community for some kind of confirmation, not one individual was surprised about this development. And while most were reluctant to go on the record, they all had heard the same story.

Careful What You Ask For: Short Term Support of Lustre a Key Issue

As Oracle exits the Lustre business, support of current Lustre installations is a real short term issue that will need to be resolved quickly.

The National Labs may be able to hold their own, but they can not provide software support to other institutions. That will be the job of the many vendors that have or support Lustre on their storage for customers to buy. Open source organizations probably have no role to perform in providing support; they can perform a role in hosting open source code repositories, bug databases, and providing an open “forum” for customers and vendors to collaborate.

The good news is that, in 2010, market forces responded to the opportunity presented by the Lustre Limbo, with new companies forming (Whamcloud) and others like Xyratex staffing up to fill in the gaps.

Norman Morse, CEO of OpenSFS, had this to say:

As your article points out, there is a huge commitment to and dependence on the Lustre file system in the Supercomputing community. Because of the requirement for a vibrant Lustre system in the future and given speculation about possible changes in Lustre support, members of the community, including some who caused Lustre to be created, founded OpenSFS specifically to ensure Lustre continues to meet community requirements and remains the preeminent open source parallel file system for high performance computing.  We are moving forward with this as our mission and we encourage all members of the Lustre community to join us – see www.OpenSFS.org.

Again as your article points out we don’t wish to comment or speculate on Oracle’s plans but OpenSFS have always stated our intention to fully cooperate with Oracle as they develop their business plan for Lustre. Through member dues, OpenSFS has resources at hand to make a major contribution to continued support and development of Lustre by funding Lustre support organizations.

The Lustre Silent Auction?

We can look forward to more Limbo for a while, but what happens next with stewardship of Lustre? Will Oracle quietly kill it like they did with OpenSolaris? Will they set the legacy code base free like they did with Grid Engine? Or will they just cash in and sell it?

Chances are that Lustre is being shopped around and we won’t hear a peep from the Dark Tower until a sale is announced. That could be months, and that kind of prolonged uncertainty would not be good for the Lustre community.

Now, if you’re wondering who might buy Lustre, you probably should look at who stands to make a lot of money supporting Lustre or selling disk storage systems that run it. I’m thinking DDN, Xyratex, or even Whamcloud. At last count, there was something like 100 Lustre engineers within Sun, so continued development is going to require deep pockets.

A Great Opportunity for the Lustre Community

As I wrote in a recent story about Lustre engineers joining Whamcloud, Oracle’s Lustre bumbling has rallied the Lustre community in a way that might not have been possible before. With the formation of HPCFSOpenSFS, and the European Open Filesystems Group, the community has done remarkable job of organizing to ensure that the popular open source file system remains viable for their pending supercomputer plans.

I know what you’re thinking; since Lustre is open source, maybe it doesn’t need a corporate holding company. The National Labs helped spawn Lustre and they can just take it back.

Why There isn’t an Easy Answer

Herein lies the rub: according to the published Lustre roadmaps, the future of Lustre on Linux is the incorporation of ZFS. Who owns that code? Oracle. So this divorce will come with strings.

Speculation and second-guessing aside, Lustre remains important for lot of reasons. Half the systems on the TOP500 run it, and there’s no open source replacement out there at the moment that is ready for prime time. And while no one in the Lustre community would probably cry if Oracle fell off a cliff, any kid can tell you that Bumbles bounce.

Comments

  1. I remain far, far more interested in other file systems like Ceph and glusterfs. My user-level experiences with Lustre have been less than favorable. It *was* the only game in town. It isn’t now.

  2. Fernanda Foertter says:

    Just when it was getting good too! It looked as if things were beginning to really mature.

  3. I wish IBM did a better job of promoting their General Parallel File System (GPFS).

  4. IBM does a great job of promoting GPFS tbf, they are also quite good at charging for it as well hence for many the interest in the open sourced lustre, gluster and Ceph filesystems remains strong.

    NB there remains strong lustre support out there, both in the community and via various contracts as mentioned in the article so no lustre is not yet dead

    Pete

  5. Hehe, not really surprised about that and not really missing it. Never liked the idea of SAN file systems like gpfs either, but with gluster for the simple cases (where fuse can do the trick) and fhgfs for HPC, I really got nothing to worry about (and also got rid of the many sleepless nights those LBUGs gave me :) )

  6. Hi guys,
    There is quite a nice alternative to Lustre -called FhGFS. It is currently not open source but free to use and does not has all the maintenance and installation problems that come with Lustre

  7. I was under the impression that btrfs was the way to go!

  8. GPFS makes some people twitchy with respect to licensing, too. The Linux version pushes the definition of derived work by digging into some interfaces. To be fair, it’s a non-technical issue that no stakeholders are interested in pursuing because of the general fuzziness, but it gave GPFS a kinda strange feeling that still lurks. CXFS kinda pushes that line a bit, too.

    A deep comparison between current versions of Lustre, Ceph, glusterfs, PVFS2, and the proprietary file systems (GPFS, FhGFS, etc.) would be quite useful. Add in pNFS with different back-ends (free, proprietary) for yet more options. There certainly is no shortage of options now.

    I don’t have the facilities or time, but I bet it’d be accepted to an academic journal or conference if someone has grad students, facilities, and enough funding lee-way… (I also bet some organizations would be willing to pay for a white paper, but I’d prefer a round of peer review.)

  9. Without Oracle, Lustre can develop more efficiently, wish Wharmcloud to give all our HPCers a good and high performance Lustre, too.

  10. Jeje, no realmente sorprendido por eso y en realidad no lo que falta. Nunca me gustó la idea de los sistemas de archivos SAN GPFS como sea, pero con Gluster para los casos simples (donde se fusionan puede hacer el truco) y fhgfs para HPC, yo realmente no tengo nada de qué preocuparse (y también se deshizo de las noches sin dormir muchas las LBUGs me dio

Trackbacks

  1. […] This post was mentioned on Twitter by insideHPC. insideHPC said: Just in: Inside Track: Oracle has Kicked Lustre to the Curb http://bit.ly/foPp0T […]

  2. […] has received an anonymous tip that Oracle has ceased development of Lustre. “We can look forward to more Limbo for a while, […]

  3. […] InsideHPC hat einen neuen Artikel über den Zustand der Lustre Entwicklung bei Oracle. Sieht wohl alles noch schlechter aus als vorher gedacht: […] Lustre engineers have reportedly been encouraged to apply for other positions within the company. […]

  4. […] next release of the Lustre file system, version 2.1. In light of the news first reported here that Oracle has ceased development of Lustre, getting the next release out has been a big concern for the Lustre […]

  5. […] the question is “Would you buy storage based on Lustre”, and it specifically points to Rich B’s article at […]

  6. […] DDN was in the news this week with the announcement that former HP fellow Jean Luc Chatelain had joined their executive team. Now the company has disclosed that they are staffing up Lustre engineering in a big way in a move to fill the gap left by Oracle. […]

Resource Links: