srb@cuci.nl writes:
>Duane Wessels <wessels@nlanr.net> wrote:
>>srb@cuci.nl writes:
>>>What happens if Squid has to swap? Are the access patterns
>>>"civilized"? Or will it thrash badly?
>
>>Probably the latter.
>
>Already feared that.
>
>>>Incidentally, are there any plans to move most of the Store Management
>>>to disk, so that Squid becomes less of a memory hog for large caches
>>>(he said, while he was contemplating expanding the proxy cache
>>>to a 4 or 6 GB striped partition)?
>
>>Yes, but nothing concrete yet.
>
>Ah. Well..., I still have about 24 hours until my cache is converted,
>by then it has to run...
>
>What's the current design policy on squid? If such a scheme would be
>made possible where most of the store management is off-loaded onto
>disk (so that the natural disk cache of the OS takes care of caching),
>would it be a requirement that the old way (most in memory) still works?
>If yes, would we settle for compile-time switchable?
>If given enough room for maneuvering, I might take a stab at
>reducing the memory requirements.
Send me code that works.
Received on Fri Jul 26 1996 - 10:06:00 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:32:42 MST