High load squid server

Hi all,

I set up this tiny... :) server to accomodate our corporation internet
access.

Sever - Now:
Caldera e-Server 2.3,
Pentium Pro 200 MMX,
with 160 Mb RAM,
4 * 4 Gb SCSI-2,
Raid 0 (I use this option just because the smartarray is already installed
and I only have one dual channel scsi controller)

Squid:
I use dlmalloc, async-io, LFUDA, quite large acl and regex configuration, 10
Gb disk for cache directory.

Recently, I notice that CPU usage is considerably high. And next month, I
want to replace this "old but nice" box, with single Xeon-866, 1 Gb RAM, two
dual channel W-U3 SCSI-controllers which is 4*18 Gb W-U3 SCSI 10K rpm disk
will be installed.

Here, I want to ask some question before I move to the new-box, first why
total memory accounted from the cachemgr isn't the same with the result by
"top"? Second, what is the meaning of Storage Replacement Treshold? Third,
are there any disk partitioning consideration regarding with squid
performance? Last, are the CPU load normal, do i need to think about dual
CPU options?

Any help would be very appreciated.

Wisnu.

ps. this is my squid report.

Squid Object Cache: Version 2.3.STABLE3

Start Time: Fri, 16 Jun 2000 08:01:56 GMT
Current Time: Mon, 26 Jun 2000 02:24:40 GMT

Connection information for squid:
        Number of clients accessing cache: 1717
        Number of HTTP requests received: 5910210
        Number of ICP messages received: 289076
        Number of ICP messages sent: 289076
        Number of queued ICP replies: 0
        Request failure ratio: 0.00%
        HTTP requests per minute: 420.3
        ICP messages per minute: 41.1
        Select loop called: 58023024 times, 14.542 ms avg
Cache information for squid:
        Request Hit Ratios: 5min: 48.9%, 60min: 39.6%
        Byte Hit Ratios: 5min: 26.1%, 60min: 28.8%
        Storage Swap size: 7623294 KB
        Storage Mem size: 32744 KB
        Storage Replacement Threshold: 0.000000
        Mean Object Size: 11.09 KB
        Requests given to unlinkd: 0
Median Service Times (seconds) 5 min 60 min:
        HTTP Requests (All): 0.17711 0.68577
        Cache Misses: 2.13280 1.81376
        Cache Hits: 0.12106 0.18699
        Near Hits: 2.25116 1.91442
        Not-Modified Replies: 0.05633 0.10857
        DNS Lookups: 1.00974 0.64769
        ICP Queries: 0.00000 0.00000
Resource usage for squid:
        UP Time: 843764.093 seconds
        CPU Time: 204409.390 seconds
        CPU Usage: 24.23%
        CPU Usage, 5 minute avg: 78.55%
        CPU Usage, 60 minute avg: 75.72%
        Maximum Resident Size: 0 KB
        Page faults with physical i/o: 2478173
Memory usage for squid via mallinfo():
        Total space in arena: 129117 KB
        Ordinary blocks: 119900 KB 149283 blks
        Small blocks: 0 KB 0 blks
        Holding blocks: 3180 KB 2 blks
        Free Small blocks: 0 KB
        Free Ordinary blocks: 9216 KB
        Total in use: 123080 KB 95%
        Total free: 9216 KB 7%
Memory accounted for:
        Total accounted: 80789 KB
File descriptor usage for squid:
        Maximum number of file descriptors: 1024
        Largest file desc currently in use: 395
        Number of file desc currently in use: 342
        Files queued for open: 0
        Available number of file descriptors: 682
        Reserved number of file descriptors: 100
        Store Disk files open: 9
Internal Data Structures:
        694517 StoreEntries
          7679 StoreEntries with MemObjects
          7581 Hot Object Cache Items
        687253 Filemap bits set
        687252 on-disk objects
Received on Sun Jun 25 2000 - 21:58:54 MDT