Hans K. Rosbach
hk at isphuset.no
Wed Jun 8 11:21:12 UTC 2011
Our company has started using glusterfs in a mail cluster
that will hopefully soon go into production phase.
So I thought I should give some feedback on what I perceive as
missing features today.
-SCTP support, this might not be a silver bullet but it feels
very strange that a cutting edge project such as glusterfs
does not support this already.
-Improved ECC suitable for jumbo frames
Features that might need glusterfs code changes:
-Multistreaming (Separate stream for metadata and stats f.ex?)
-Multihoming (failover when one nic dies)
-Documentation for how to add new translators and how to change
translator options in 3.1/3.2 is a bit poor. I first tried to
change the config files manually but failed miserably before
I found a description on the mailinglist of what commands to use.
-Ability to have the storage nodes autosync themselves.
In our setup the normal nodes have 2x1Gbit connections while the
storage boxes have 2x10Gbit connections, so having the storage
boxes use their own bandwidth and resources to sync would be nice.
And I don't really like the "read all files to sync them" way of
doing things, it should be possible to do this by communicating
only metadata between the servers. Reading many many millions of
tiny files could take ages.
-An ability for the clients to subscribe to metadata updates for
a specific directory would also be nice, so that it can cache that
folders stats while working there and still know that it will not
miss any changes. This would perhaps increase overhead in large
clusters but could improve performance by a lot in clusters where
several nodes work in the same folder (mail spool folder for example).
Manual static subscriptions would be nice, but automatic temporary
subscriptions would also be a nice addition (but still the ability
to force specific folders manuelly).
Hopefully some of this will seem interesting to others aswell.
Hans Kristian Rosbach
ISPHuset Nordic AS
More information about the Gluster-devel