Re: --enable-async-io code status

From: Jens-S. Voeckler <voeckler@dont-contact.us>
Date: Mon, 4 Jan 1999 10:38:54 +0100

On Wed, 30 Dec 1998, David J N Begley wrote:

]> >Platforms for which it could be considered non-experimental is probably
]> >Linux with glibc-2.0.7 and Solaris 2.6?
]>
]> Probably, I don't know about Solaris.
]
]What would you like people to look for in reporting whether or not
]"enable-async-io" works on Solaris? (FWIW, I'm using this option on Solaris
]2.5.1.)

I wouldn't know the concrete metrics for that, but "it works" (e.g. the
cache answers fast and reliably) with my local test caches (2.5.1+2.6, low
load), but "it does not work" (e.g. the cache slowed considerably compared
to non-aio) when testing on one of my production caches (2.5.1, toplevel
cache). The latter may be a further indication that my disk subsystem on
the old caches is not performing as well as I need it to.

One further "quirk": When 'blasting' a squid with (all MISS) requests, and
the disk subsystem being a RAM disk, the nonthreaded version was
performing better (2.6; compared to threaded versions with 40 and 80
thread on the same RAM disk amount).

Le deagh dhrachd,
Dipl.-Ing. Jens-S. Vckler (voeckler@rvs.uni-hannover.de)
Institute for Computer Networks and Distributed Systems
University of Hanover, Germany; +49 511 762 4726
Received on Mon Jan 04 1999 - 02:36:55 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:43:54 MST