On Wed, 2007-08-22 at 12:02 +0900, nac wrote: > Hi all, > now I'm using JFS on Gentoo linux (Kernel 2.6.19) box. > This box is really simple (no great cpu, no RAID) just for > testing. I'm curious about memory usage of JFS. > > This box has 512 MB RAM and uses less than 20 MB ram before many > disk access, like find command for /usr/portage. > But after that, memory utilization has fairly increased. > In slabtop, jfs_ip seems using memory. > > The portage directory has a lots of small files, its size is at most 10kb and > average file size is much more smaller than 10kb. > And number of files is over 120000. > > > <<before "find /usr/porage">> > $ free > total used free shared buffers cached > Mem: 491912 30536 461376 0 24 16076 > -/+ buffers/cache: 14436 477476 > Swap: 1004052 0 1004052 > > > <<after "find /usr/portage">> > $ free > total used free shared buffers cached > Mem: 491912 348784 143128 0 24 159220 > -/+ buffers/cache: 189540 302372 > Swap: 1004052 0 1004052 > > $ cat /proc/slabinfo > slabinfo - version: 2.1 > # name <active_objs> <num_objs> <objsize> <objperslab> > <pagesperslab> : tunables <limit> <batchcount> <sharedfactor> : slabdata > <active_slabs> <num_slabs> <sharedavail> > <<<snip>>> > jfs_mp 36275 36305 72 53 1 : tunables 120 60 8 : > slabdata 685 685 0 > jfs_ip 151376 151376 876 4 1 : tunables 54 27 8 : > slabdata 37844 37844 0 > udf_inode_cache 0 0 380 10 1 : tunables 54 27 8 : > slabdata 0 0 0 > <<<snip>>> > > JFS(jfs_ip) memory utilization is not small, I think. > Is this memory utilization is normal?
Yes this is normal. jfs has a larger in-memory inode than most file systems, but I'm not aware that it causes any real problems. > And memory utilization stays at the level, so I don't think memory > leaking, this is correct? Yes, it's correct. Applications that stat a lot of files without reading any data from them, like find and updatedb, will cause a lot of file data to be cached in the inode cache, and directory cache (dcache). These caches will shrink when the memory is needed for something else, but the linux kernel won't try to free cached data unnecessarily. If you have issues with high inode cache and dcache usage causing useful data to be swapped out of memory, you may be interested in tuning /proc/sys/vm/vfs_cache_pressure. A high number here (could be in the thousands) will cause the kernel to try to free inode cache and dcache more aggressively. > $ cat /proc/meminfo > MemTotal: 491912 kB > MemFree: 141732 kB > Buffers: 24 kB > Cached: 160136 kB > SwapCached: 0 kB > Active: 149048 kB > Inactive: 17432 kB > HighTotal: 0 kB > HighFree: 0 kB > LowTotal: 491912 kB > LowFree: 141732 kB > SwapTotal: 1004052 kB > SwapFree: 1004052 kB > Dirty: 4 kB > Writeback: 0 kB > AnonPages: 6340 kB > Mapped: 5052 kB > Slab: 178424 kB > SReclaimable: 172260 kB > SUnreclaim: 6164 kB > PageTables: 292 kB > NFS_Unstable: 0 kB > Bounce: 0 kB > CommitLimit: 1250008 kB > Committed_AS: 57844 kB > VmallocTotal: 532472 kB > VmallocUsed: 6824 kB > VmallocChunk: 525060 kB > > > $ cat /proc/version > Linux version 2.6.19-gentoo-r5 ([EMAIL PROTECTED]) (gcc バージョン 4.1.1 (Gentoo > 4.1.1)) #1 SMP Tue Mar 6 12:52:45 JST 2007 > > ------- > nac. > [EMAIL PROTECTED] -- David Kleikamp IBM Linux Technology Center ------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/ _______________________________________________ Jfs-discussion mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/jfs-discussion
