Re: [squid-users] Re: Squid limits and hardware spec

2004-12-03 Thread Michael Pophal
As mentioned before, try the new version of calamaris (v2.99). There is
a new report included, which shows you in the 'Requested extensions'
report some nice information about object freshness:
- ratio fresh/stale 
- ratio unmod/mod
This helps you to improve your squid refresh_pattern.

Michael

On Thu, 2004-12-02 at 22:39, Adam Aube wrote:
 Martin Marji Cermak wrote:
 
  I have been playing with Squid under a heavy load and there are some
  stats. I am trying to maximise the Byte Hit Ratio value. I got 13%
  average, but I am not happy about this number - I want it higher
 
 To increase your byte hit ratio, you can:
 
 1) Switch to one of the heap cache replacement policies
 2) Tune your refresh_pattern settings to make Squid cache more aggressively
 
 See the FAQ and default squid.conf for details on these items.
 
 However, before going through the tuning, run an analysis tool (such as
 Calamaris) on your logs to see what your traffic pattern is like. This will
 show you what a reasonable byte hit ratio would be.
 
 If, for example, 70% of your traffic is dynamic content (which usually
 cannot be cached), then a 13% byte hit ratio is actually pretty good.
 
  USED HARDWARE:
  Processor: P4 1.8GHz
  Memory:1 GB
  Hardisk:   40 GB IDE 7200rpm
 
  Requests: 180 req/sec (peak), 60 req/sec (day average).
 
 According to posts from Squid developers, a single caching Squid box has an
 upper limit of about 300 - 400 requests/second. This isn't too bad,
 considering you are using a single IDE disk for the entire system.
 
  maximum_object_size 51200 KB (SHOULD I MAKE IT HIGHER ???)
 
 Actually, you might want to make it lower. Most web requests will not be for
 50 MB files, and your byte hit ratio will be hurt if a 50 MB file that is
 requested once forces out fifty 1 MB files that are accessed twice each.
 
 The default is generally acceptable, unless log analysis shows large numbers
 of requests for larger files.
 
  cache_dir aufs /cache 25000 16 256
 
 You should size your cache to hold about a week's worth of traffic. Just
 watch your memory usage (1 GB of cache ~ 10 MB of memory for metadata).
 
  cache_mem 8 MB
 
 This is generally fine - the OS will generally use free memory to cache
 files anyway, which will have the same effect as boosting this setting.
 
  I am going to install a new box with SCSI disks so I will report to you
  how the performance will change.
 
 Best disk performance will be achieved with multiple small, fast SCSI disks
 dedicated to Squid's cache, each with its own cache_dir (no RAID), and
 round-robin between the cache_dirs.
 
 Adam




[squid-users] Re: Squid limits and hardware spec

2004-12-02 Thread Adam Aube
Martin Marji Cermak wrote:

 I have been playing with Squid under a heavy load and there are some
 stats. I am trying to maximise the Byte Hit Ratio value. I got 13%
 average, but I am not happy about this number - I want it higher

To increase your byte hit ratio, you can:

1) Switch to one of the heap cache replacement policies
2) Tune your refresh_pattern settings to make Squid cache more aggressively

See the FAQ and default squid.conf for details on these items.

However, before going through the tuning, run an analysis tool (such as
Calamaris) on your logs to see what your traffic pattern is like. This will
show you what a reasonable byte hit ratio would be.

If, for example, 70% of your traffic is dynamic content (which usually
cannot be cached), then a 13% byte hit ratio is actually pretty good.

 USED HARDWARE:
 Processor: P4 1.8GHz
 Memory:1 GB
 Hardisk:   40 GB IDE 7200rpm

 Requests: 180 req/sec (peak), 60 req/sec (day average).

According to posts from Squid developers, a single caching Squid box has an
upper limit of about 300 - 400 requests/second. This isn't too bad,
considering you are using a single IDE disk for the entire system.

 maximum_object_size 51200 KB (SHOULD I MAKE IT HIGHER ???)

Actually, you might want to make it lower. Most web requests will not be for
50 MB files, and your byte hit ratio will be hurt if a 50 MB file that is
requested once forces out fifty 1 MB files that are accessed twice each.

The default is generally acceptable, unless log analysis shows large numbers
of requests for larger files.

 cache_dir aufs /cache 25000 16 256

You should size your cache to hold about a week's worth of traffic. Just
watch your memory usage (1 GB of cache ~ 10 MB of memory for metadata).

 cache_mem 8 MB

This is generally fine - the OS will generally use free memory to cache
files anyway, which will have the same effect as boosting this setting.

 I am going to install a new box with SCSI disks so I will report to you
 how the performance will change.

Best disk performance will be achieved with multiple small, fast SCSI disks
dedicated to Squid's cache, each with its own cache_dir (no RAID), and
round-robin between the cache_dirs.

Adam