Re: [SQU] high-availability proxies

Hi,

On Fri, 22 Dec 2000, Florin Andrei wrote:
> >http://wwwcache.ja.net/JanetService/PilotService.html
>
> Please, can you detail a little bit on this?
> - what does it mean "national cache servers"?

JWCS provide a top-level cache service for UK academia - the national
cache servers are the machines that run this service. Below these,
individual sites run cache servers for their local user communities.

> - how many real caches do you have?

At last count, we had about 40 machines spread across three nodes, with
a combined capacity of going on for 2TB.

> - how many reqs per second do you serve?

This, and other such information, is available from our stats pages on the
site (http://wwwcache.ja.net/stats/MRTG/) - on one of our more heavily
weighted machines (wergilb) the maximum reqs/min from the past few weeks
was 10,900 - so about 181 reqs/sec peak on one machine.

> - what is the average traffic through the balancer?

We have a potential 100Mbps per node - bandwidth utilisation information
can be found at http://wwwcache.ja.net/stats/MRTG/uplinks.html

Over the past year, we think our peak data shippage was in the region of
800GB/day. Recent figures are low as we've just entered the vacation
period :-)

> - how many client machines connect to the balancer?

Difficult to say - we service about 170 UK HE institutions, each of which
will have their own local clients (who either use a local cache which
parents onto us, or who will go directly to the JWCS). We have also seen
recently the addition of FE institutions to the JANET network, so there is
potential for some or all of these c.400 establishments to use the cache
service.

I hope this answers some of your queries - if you have any further queries
or comments, please let us know.

Regards,

-- 
Andrew Veitch, National & Local Web Cache Support    andrew.veitch@man.ac.uk
Manchester Computing, University of Manchester       http://wwwcache.ja.net/
--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html