I did some testing of a couple of x4500 servers to see how well they scaled in terms of number of filesystems. Both were running snv_97 with the following config:
# zpool create -f export mirror c0t0d0 c1t0d0 mirror c4t0d0 c6t0d0 mirror c0t1d0 c1t1d0 mirror c4t1d0 c5t1d0 mirror c6t1d0 c7t1d0 mirror c0t2d0 c1t2d0 mirror c4t2d0 c5t2d0 mirror c6t2d0 c7t2d0 mirror c0t3d0 c1t3d0 mirror c4t3d0 c5t3d0 mirror c6t3d0 c7t3d0 mirror c0t4d0 c1t4d0 mirror c4t4d0 c6t4d0 mirror c0t5d0 c1t5d0 mirror c4t5d0 c5t5d0 mirror c6t5d0 c7t5d0 mirror c0t6d0 c1t6d0 mirror c4t6d0 c5t6d0 mirror c6t6d0 c7t6d0 mirror c0t7d0 c1t7d0 mirror c4t7d0 c5t7d0 mirror c6t7d0 c7t7d0 spare c7t0d0 c7t4d0 # zfs set devices=off export # zfs set setuid=off export # zfs set aclmode=passthrough export # zfs set aclinherit=passthrough export # zfs create export/user On only one of them I turned on NFS: # zfs set sharenfs=nosuid,sec=krb5i:krb5p export On each of them I then created 6000 filesystems, 1000 at a time, measuring how long it took to create each batch of 1000, and how long the reboot cycle was for increasing filesystem counts. No NFS: #fs,create time,reboot time 1000,369,471 2000,1063,268 3000,2340,318 4000,4622,570 5000,6918,832 6000,9416,960 NFS: #fs,create time,reboot time 1000,670,492 2000,2981,446 3000,7556,934 4000,15423,1406 5000,25222,2035 6000,38322,2899 Performance started out great, taking less than .4 seconds on average per filesystem, and less than .7 when shared, with a reboot cycle of only 7-8 minutes. Once 6000 filesystems were created, things were a *lot* slower, taking almost 10 seconds to create a new filesystem, and almost 40 seconds if sharing was enabled. I could live with a 16 minute boot cycle, but with NFS enabled it was over 40 minutes. I ran a similar test way back under U4: #FS create/avg 1000 589/.59 2000 709/.71 3000 783/.78 4000 836/.84 5000 968/.97 6000 930/.93 7000 961/.96 8000 1045/1.05 9000 1098/1.10 10000 1165/1.17 Creation time was actually considerably better under U4 than snv_97. I was playing with SXCE to get a feel for the soon to be released U6. Performance wise, I'm hoping U6 will be better, hopefully some new code in SXCE was introduced that hasn't quite been optimized yet... Last date I heard was Nov 10, if I'm lucky I'll be able to start testing with actual U6 soon. -- Paul B. Henson | (909) 979-6361 | http://www.csupomona.edu/~henson/ Operating Systems and Network Analyst | [EMAIL PROTECTED] California State Polytechnic University | Pomona CA 91768 _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
