[nycbug-talk] glusterfs

[Miles Nordin]

>>>>> "jc" == Jesse Callaway <bonsaime at gmail.com> writes:

    jc> Miles, I'm sure you have something smarmy and intelligent to
    jc> say about glusterfs

I don't.

I'm not using it yet is the problem.

I've been meaning to use it at work.  I don't have any problem with it
if that's what you mean.  To me this type of thing looks like the
future---not necessarily the fuse/rump side of it but the
disk-[optionalredundancy]-filesystem-redundancy-filesystem layering,
and secondly the idea that the storage backplane needs to be inside a
network switch and not along a single link of any kind, not even if
it's a single link of FC-SW.  I like the pluggable back-ends and the
ghetto-HSM policy stuff.  I like the way they can supposedly lose
whole chunks of unredundant storage bricks without completely shitting
themselves, just losing some files or subdirectories.

but I have not tried it so...if you find it doesn't work that's kind
of a big negative point.

    jc> Sun has a cool project called Celeste going on

I will look at this!

I was impressed that Lustre has a plausible, rigid timetable.  However
it seems like ZFS's super-efficient snapshots (in space consumption
and in creation/deletion time, more efficient than vmware or oracle
db) are neither in Lustre nor on the Lustre timetable, which is a big
loss.  I'm not sure the efficient way to cram that back into the new
model.  probably need to dig into it more.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 304 bytes
Desc: not available
URL: <https://lists.nycbug.org:8443/pipermail/talk/attachments/20090503/67086dca/attachment.bin>