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

Digging in to Oracle's Lustre strategy: less is less

I’ve just been through the Lustre strategy presentation that Peter Bojanic briefed last week (previously linked here), and had a great conversation with one of my more knowledgeable friends on what look like changes in the Lustre strategy as the product moves from Sun to Oracle. I’d say the strategy is to lock up future innovation in Lustre to drive the purchase of Sun/Oracle hardware. I’d further guess that if that move does not result in increased business, Oracle will shed the Lustre effort entirely, leaving it to the community in whatever crippled state the open source portion of 2.0 is in at the time. But I’m getting ahead of myself.

Oracle logoFor Lustre 1.8 users with existing support contracts will be “allowed” to renew them (slide 13, bullet 3), but there is no supported migration path for users on storage hardware that is not made by Sun/Oracle for customers wanting to upgrade from Lustre 1.8 to 2.0 (slide 13, bullet 2). Lustre 2.0 will only be supported and qualified (by Sun/Oracle) on Sun/Oracle hardware.

Customers who want the advanced features of Lustre 2 can purchase Oracle’s integrated storage configurations to get the highest levels of reliability, performance, and support (slide 13, bullet 4)

There also won’t be any support available from Sun/Oracle for the software-only version of Lustre 2.0 (slide 9, bullet 3), and it looks like they are planning to add features that are value add only for Sun/Oracle hardware customers (slide 5, bullet 2) that “may or may not” be open source. I’d read that as “will not.”

Comments

  1. Amusingly, I wrote about this very topic *exactly* one year ago – http://pl.atyp.us/wordpress/?p=2047. Long story short, it’s not unlikely that Lustre will revert to the community – probably led by LLNL/ORNL or similar – and that might very well be the best thing for it.

  2. As I commented on Joe’s blog about Lustre they have committed to keeping the canonical release branch of the filesystem in their public repos and to keep it GPL (they know very well they can’t distribute non-GPL kernel code from their comments on Linux and ZFS in that same presentation), but there may well be user space bits that they may keep closed source.

    The thing I wonder about is whether they would consider supporting btrfs (their own pet FS project) as an alternative backend for Lustre (once btrfs has stabilised), that should give them the error checking they currently want from ZFS (and can’t get under Linux currently).

Trackbacks

  1. [...] seeing the slides, speaking with some of the support team, seeing John West and John Leidel’s discussion of Lustre 2.0 on InsideHPC [...]

  2. [...] example, that Oracle today couldn’t (or wouldn’t) build Ranger for TACC. This plus the recent Lustre news certainly adds fuel to this community’s speculation that Oracle’s commitment to HPC [...]

  3. [...] an open letter to customers on their website, outlining their take on how Oracle’s recent shifts in Lustre policy will impact Terascala customers Recently, Oracle has announced a planned roadmap for the Lustre [...]

Resource Links: