On 8/28/06, Adrian Chadd <adrian@xxxxxxxxxxxxxxx> wrote:
On Mon, Aug 28, 2006, Pranav Desai wrote:
> >A single drive and you expect 1000 req/s? Is this a solid-state drive or
> >something with infinitely small seek time?
> >
> No, its just a regular drive. I am already working on that, but its
> kind of difficult to get things changed quickly :-).
> But the good part is that even with 1 disk its able to do 1000 req/s
> atleast for the first 6 hrs, so I am sure I can push it more once I
> get some better hardware.
I think you'll find that the performance is degrading because:
* (a) the disk write queues are slowly filling up and taking longer to write;
* (b) the disk is filling up and fragmentation+object replacement+non-linear
file allocation kicks in.
Faster hardware won't fix that. More disks may. COSS will help a lot for
the smaller objects. But a lot more work needs to be done to improve the
Squid disk store under load.
for (a), shouldnt the disk queue also affect the performance during
the first phase of the test. The first phase is almost perfect with no
degradation in times.
If I may send you the graphs, in which it seems very wrong that the
degradation happens so suddenly.
Maybe I will try a few test with a longer first phase to see if the
disk queue catches up and degrades the performance.
I will be trying COSS shortly. Even I am expecting it to do much
better. What other kind of work do you have in mind. Another
filesystem or something ?
I am thinking of giving this http://logfs.sourceforge.net/ a try.
But I dont know if it needs changes in squid to make efficient use of
the filesystem.
> It didnt in the first phase. In fact it did very well in the first
> phase compared to the first phase of 2.5, but somehow after the idle
> phase it just wasnt able to recover and the polybench wasnt even able
> to push to 1000 req/s because the response times were so high.
Whats your squid.conf look like? It sounds like the disks just couldn't
keep up with your request rate.
visible_hostname 10.51.6.102
cache_dir aufs /var/cache/squid 30720 16 256
http_port 8080
request_body_max_size 0
snmp_port 3401
negative_ttl 0 minutes
pid_filename /var/run/squid.pid
coredump_dir /var/log/squid
cache_effective_user squid
cache_effective_group squid
cache_access_log /var/log/squid/access.log
cache_log /var/log/squid/cache.log
cache_store_log none
cache_swap_log /var/log/squid/swap.log
logfile_rotate 10
icp_port 3130
icp_query_timeout 20000
log_icp_queries on
extension_methods SEARCH PROPPATCH
forwarded_for on
acl all src 0.0.0.0/0.0.0.0
acl localhost src 127.0.0.1 10.51.6.102
acl manager proto cache_object
acl snmppublic snmp_community public
http_access allow localhost
miss_access allow all
http_access allow all
snmp_access allow snmppublic all
memory_pools off
Adrian
--
------------------------------
http://pd.dnsalias.org
