On Mon, 15 Feb 1999, Henrik Nordstrom wrote:
]> Going with what you've said, I'm actually still not that suprised that
]> I saw a performance boost when I essentially disabled the OS level
]> buffer cache on the cache filesystems and upped
]> cache_mem.
]
]I am suprised. But then, perhaps the only objects worth storing in
]memory are small ones which changes very often.
Hmm, I didn't see any difference for MISSes when I activated the Solaris
"forcedirectio" mount option (since 2.6) which is said to transfer from
user memory to io without a kernel copy. But then, (a) it was a lab
environment and (b) I didn't look at the effects on HITs, because I was
banging an empty Squid.
Judging by what you explained on cache_mem, I should *not* use that mount
option, because the kernel buffer cache is actually my object store (for
larger objects)?
Le deagh dh�rachd,
Dipl.-Ing. Jens-S. V�ckler (voeckler@rvs.uni-hannover.de)
Institute for Computer Networks and Distributed Systems
University of Hanover, Germany; +49 511 762 4726
Received on Mon Feb 15 1999 - 03:23:13 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:44:33 MST