[squid-users] Peak Load on single instance of squid running on Dual Xeon 2.2GHz

From: Venkatesh. K <[email protected]>
Date: Wed, 17 Sep 2003 21:12:29 +0530

Hi,

I am managing a busy cache with 1000 active users. There server is Dual Xeon
2.2 GHz / 2304MB RAM and 6x36GB 10K RPM. Currently, I am using 2 disks for
cache.

I am using 2.5.STABLE3 with aufs/reiserfs.

The CPU utilization is peaking at 95% and above when the box does about 150+
req/sec.

Can this box handle more traffic ?

Here are the stats

sample_start_time = 1063811461.622799 (Wed, 17 Sep 2003 15:11:01 GMT)
sample_end_time = 1063811761.643423 (Wed, 17 Sep 2003 15:16:01 GMT)
client_http.requests = 146.689916/sec
client_http.hits = 73.458283/sec
client_http.errors = 0.000000/sec
client_http.kbytes_in = 125.878013/sec
client_http.kbytes_out = 801.084928/sec
client_http.all_median_svc_time = 0.134979 seconds
client_http.miss_median_svc_time = 1.461314 seconds
client_http.nm_median_svc_time = 0.009189 seconds
client_http.nh_median_svc_time = 1.461314 seconds
client_http.hit_median_svc_time = 0.016481 seconds
dns.median_svc_time = 0.163043 seconds
unlink.requests = 0.000000/sec
page_faults = 0.379974/sec
select_loops = 107.452613/sec
select_fds = 951.731238/sec
average_select_fd_period = 0.001049/fd
median_select_fds = 6.972656
swap.outs = 18.715380/sec
swap.ins = 87.477320/sec
swap.files_cleaned = 0.000000/sec
aborted_requests = 9.106041/sec
syscalls.polls = 151.242936/sec
syscalls.disk.opens = 0.000000/sec
syscalls.disk.closes = 0.000000/sec
syscalls.disk.reads = 0.000000/sec
syscalls.disk.writes = 99.966461/sec
syscalls.disk.seeks = 0.000000/sec
syscalls.disk.unlinks = 0.000000/sec
syscalls.sock.accepts = 97.243315/sec
syscalls.sock.sockets = 57.489381/sec
syscalls.sock.connects = 56.899422/sec
syscalls.sock.binds = 56.959418/sec
syscalls.sock.closes = 120.631707/sec
syscalls.sock.reads = 516.214512/sec
syscalls.sock.writes = 451.912266/sec
syscalls.sock.recvfroms = 10.835922/sec
syscalls.sock.sendtos = 6.059583/sec
cpu_time = 292.290000 seconds
wall_time = 300.020624 seconds
cpu_usage = 97.423302%
Received on Wed Sep 17 2003 - 09:37:39 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:19:55 MST