Hi, On Wed, 2011-11-09 at 09:57 -0500, Nicolas Ross wrote: > (...) > > > On some services, there are document directories that are huge, not that > > much in size (about 35 gigs), but in number of files, around one million. > > One service even has 3 data directories with that many files each. > > (...) > > We're leaning towards changing the file system for those document > directories (agian hierachy of, not 1 mil files per directory ;-) to ext4. > > One or two question pops in my mind. > > 1. Can a node mount the volume read-write, while another one mount it > read-only ? > No (assuming you mean ext4 here) > 2. I curently experience intermitent high i/o contention on the volume > containing this data, the website still responds fase, but images takes more > time to load when it arrives. I highly suspects, that since it's a ton of > files spread trough out the disk, the it's the added latency of GFS that > causes my problem. > > So, how will perform an ext4 files over fiber versus a local one ? > ext4 should be just as fast, or even faster depending on what storage it is running on. The issue though is whether you need to share the data or not? GFS2 is slower in certain cases because of the overheads of the keeping everything cluster coherent. If you don't need cluster coherency, then maybe it is not the right solution, and ext4 may be preferable, Steve. -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
