Re: some linux tuning

Brian writes:
> Just looking for some basic linux/squid tuning tips. What I have done so
> far:
>
> Background on machine:
>
> PII450
> 512MB SDRAM
> DPT3334UW RAID controller doing RAID0 with 64MB disk cache
> Linux 2.2.7ac4
>
> We have about 1300 dialup lines, so about 1300 people max could be going
> thru box at a time. (usually its more like 400-500). Some clients on
> ISDN/T1's/ethernet/etc.

  How much bandwidth to the 'net do you have? I think that's one of
the critical factors in how much performance you need to get out of the
cache. Better still is if you can estimate peak "down-stream"
bandwidth to your users. Do you have monitoring software on your
network so you could add up total bandwidth flows to each terminal
server, each outgoing dedicated line, etc. to get a peak "incoming
bandwidth" flow?

  Can't help on the Linux fs param tuning - that's still a bit alien to
an old BSD bigot...

[...]
> 2. File descriptors increased to 8192 (2.2.7ac4)
> 3. Hardware RAID0 with 64MB disk cache (dpt3334uw controller)
> 4. 512MB memory.........cache_mem 160 MB (is that a good value? 1/3?)
> 5. dns_children 10
> 6. squid 2.2STABLE3

  These are similar guesses to what I've been making as to general
hardware to maximize performance.

> 7. ./configure --prefix=/usr/local/squid --enable-async-io
> --disable-ident-lookups --enable-gnuregex
> 8. Foundry ServerIron redirecting requests to squid
> 9. cache_dir /usr/local/squid/cache 13500 16 256
> Is that a little small? Do you think more space is needed?

Gut level: maybe a little small, without doing the math. There's a
paper by Peter Danzig (originally of the Harvest project, now at
NetCache) on estimating a good cache size at
  <http://www.skycache.com/sizingcaches.pdf>
I plan to do the math based on that, and go with what he says.

> 10. Satellite cache prepopulation feed (www.skycache.com)

Here's why I really wanted to reply:

  Are you actively using this now? Is it working well for you?

  We are on the brink of signing up with them; one of their tech guys
is from Hawaii, formerly with one of our rival ISPs, and he's a very
sharp guy, though he's mainly working on their Usenet service. My math
says that even just pulling news off their service would save us money,
and if we get a better cache hit rate, that's "pure gravy". But that's
only if it really works consistently in practice. Any estimate on how
much cache hits they're delivering to you, or how much bandwidth
they're successfully offloading?

> 11. Caching only nameserver ran on squid host

  Which nameserver?

> Some things I have found necessary to keep user complaints down and deal
> with non compliant sites: (are their any other tips?)
>
> quick_abort_min 0 KB
> quick_abort_max 0 KB
> quick_abort_pct 100
> half_closed_clients off
> uri_whitespace allow

  Not yet. Given the number of badly done web servers out there, I'm
amazed how well Squid deals with them so far; but we haven't put it
into full customer beta yet. We're very conservative about that kind
of thing.

  -- Clifton

-- 
 Clifton Royston  --  LavaNet Systems Architect --  cliftonr@lava.net
        "An absolute monarch would be absolutely wise and good.  
           But no man is strong enough to have no interest.  
             Therefore the best king would be Pure Chance.  
              It is Pure Chance that rules the Universe; 
          therefore, and only therefore, life is good." - AC