Re: Large, busy, caches... problems?

On Wed 26 Mar, 1997, Ian Cooper <ihc@ukc.ac.uk> wrote:
>I'm still struggling to get one of the HENSA Cache boxes to run Squid
>(1.1.8) in an efficient manner.

>Would other users of large, busy, caches care to comment on their
>experiences? (Particularly with respect to the problems encountered
>when paging or operating during the clean cycle.)

>cache_mem = 4
>cache_swap = 2000
>
>The disk cache is distributed over 8 XFS disk partitions (each
>partition on a separate physical disk)
>
>The machine in an SGI Challenge S, IP22, running IRIX 5.3 and is
>dedicated to running the cache.

here's my sharing of info:

With one of my SGI Challenge S's (our really slow one - 133MHz)/
256MB RAM (Irix 6.2) I've been experimenting with Squid 1.1.8.

cache_mem 128
cache_swap 6000
the low/high as per defaults.
max object size is 6MB.

the only other tunables i've set, as yet, are:

quick_abort 20 100 90
positive_dns_ttl 1
negative_dns_ttl 2

After 7 days uptime, it's reached 217MB process size, so little paging
and thus it runs ok.

Cache swap space is across 6 devices (2 controllers), actually 12GB
of disk space but I've set it to half and it's keeping within it.

Also, trying out Cord Beermann's calamaris.pl (which I like
already), I note:

# Request peak
Protocol sec minute peak begins at hour peak begins at
-------- ----- ------ ------------------------ ------- ------------------------
UDP 89 695 Tue Mar 25 21:33:08 1997 31033 Tue Mar 25 21:18:58 1997
TCP 41 854 Tue Mar 25 21:33:22 1997 40881 Tue Mar 25 21:08:09 1997
-------- ----- ------ ------------------------ ------- ------------------------
ALL 89 1528 Tue Mar 25 21:33:08 1997 71789 Tue Mar 25 21:08:14 1997

about yesterday's traffic. (Not sure about those per second/minute figures
though, they seem to be off by an order of magnitude from what I'd expect).

I wasn't around at that time to pay particular note to how the machine
was performing though...

for reference, here's the other summary figures for this machine:

# UDP-Request State Request % kByte % msec KB/sec
------------------------------------ ------- ------ -------- ------ ---- ------
HIT 34792 8.91 116340 82.94 6.12 545.55
 UDP_HIT_OBJ 28833 7.38 115950 82.66 6.58 610.94
 UDP_HIT 5959 1.53 389 0.28 3.93 16.62
MISS 355713 91.09 23934 17.06 4.07 16.52
 UDP_MISS 355713 91.09 23934 17.06 4.07 16.52
------------------------------------ ------- ------ -------- ------ ---- ------
Sum 390505 140274 4.25 84.41

# TCP-Request State Request % kByte % sec KB/sec
------------------------------------ ------- ------ -------- ------ ---- ------
HIT 198870 37.05 1201822 24.36 1.73 3.47
 TCP_HIT 111454 20.76 955264 20.16 1.72 4.97
 TCP_IMS_HIT 57050 10.63 22836 0.48 0.69 0.58
 TCP_REFRESH_HIT 29698 5.53 218814 4.62 2.82 2.61
 TCP_REF_FAIL_HIT 668 0.12 4907 0.10 45.2 0.16
MISS 324018 60.36 3428676 72.37 5.72 1.85
 TCP_MISS 287966 53.65 3245137 68.50 5.83 1.93
 TCP_CLIENT_REFRESH 19487 3.63 74096 1.56 3.09 1.23
 TCP_REFRESH_MISS 16506 3.08 109406 2.31 6.92 0.96
 TCP_IMS_MISS 59 0.01 36 0.00 1.22 0.50
ERROR 13885 2.59 106992 2.26 22.2 0.35
 ERR_CLIENT_ABORT 7698 1.43 90328 1.91 22.6 0.52
 ERR_CONNECT_FAIL 3322 0.62 3144 0.07 27.2 0.03
 ERR_DNS_FAIL 2176 0.41 2049 0.04 2.02 0.47
 TCP_IMS_MISS 199 0.04 0 0.00 0.86 0.00
 TCP_DENIED 188 0.04 102 0.00 0.32 1.67
 ERR_ZERO_SIZE_OBJECT 115 0.02 106 0.00 35.4 0.03
 ERR_INVALID_URL 68 0.01 43 0.00 0.75 0.85
 ERR_READ_ERROR 64 0.01 63 0.00 32.2 0.03
 ERR_READ_TIMEOUT 34 0.01 10623 0.22 949. 0.33
 ERR_INVALID_REQ 14 0.00 8 0.00 1.42 0.44
 ERR_UNSUP_REQ 5 0.00 3 0.00 0.57 1.19
 ERR_LIFETIME_EXP 2 0.00 518 0.01 660. 0.39
------------------------------------ ------- ------ -------- ------ ---- ------
Sum 536773 4737492 4.67 1.89

-- jrg.
Received on Wed Mar 26 1997 - 09:56:32 MST