NFS Performance Tuning for HP-UX 11.0 and 11i Systems
nfs performance tuning for hp-ux 11.0 and 11i systems page 81
Notes:
Page 81July 22, 2002
Copyright 2002 Hewlett- Packard Company
cachefs
CacheFS Limitations (part 2)
• Only certain files survive a CacheFS unmount or reboot
• Any file marked “un-cacheable” will be removed
Ø Writing to a cached file marks the file “un-cacheable”
Ø When a cache reaches its configured disk space or inode usage
thresholds, the LRU algorithm will select files to remove.
Ø Every cached file is represented by a 32-slot allocation map
data structure, where each slot represents a non-contiguous
chunk of the file. Any file that requires more than 32 non-
contiguous chunks to be loaded is considered “un-cacheable.”
There are three main reasons CacheFS will disable caching for a file:
• When a process on the NFS client writes to the cached file
• When a cache reaches its configured disk space or inode usage thresholds, the
LRU algorithm will select files to remove from the cache
• When 33 or more non-contiguous chunks of the cached file are referenced
Once a file is flagged as “un-cacheable,” any read requests for this file must be
serviced by the back (NFS) filesystem, effectively nullifying any CacheFS benefits
for this file. All files that have been marked “un-cacheable” are removed from the
cache directory when the CacheFS filesystem is unmounted or the NFS client system
is rebooted.