Re: [squid-users] Ideal cache placement (was Re: Why Squid is great (was: fourth cache off??))

Jon Kay wrote:
> OK, let's take a look front and center at these wondrous central LANs
> of yours. They can often be 3,4, even 5 bridge/router hops away,
> plus 3-5ms assorted basic roundtrip routing and propagation delay,
> which we see once for each part of the dual TCP handshake. To that we
> have to add 3-30ms store/forward delay for the average 12k object on an a
> assortment of 10BT and 100BT Ethernets.... Just one old Ethernet, and
> an unaggressively designed network, and you have just doubled that hit
> time of yours.

I decided to measure this number to see what I'd get between
my home and a machine at the office.

My home machine is connected to a 100 Mb/s hub.
The hub is connected to the home end of an ADSL modem pair.
That modem is connected to the other end of the pair at the
office (uplink/downlink of 256kb and 1.5Mb, I believe).
The office modem is connected to the DSL switch, which is
connected to the backbone switch, which is then connected
to the office switch, all of which are switched fast ethernet.
The link to my office is 10 Mb/sec, and then there's another
100 Mb/sec hub to which my desktop machine is connected.

Ping reports the average round-trip time to be < 7ms on this
"unaggressively designed network".

64 packets transmitted, 64 packets received, 0% packet loss
round-trip min/avg/max/stddev = 6.555/6.775/12.308/0.710 ms

That's not bad for 7 intermediate devices between two machines,
and is hardly a death knell for centralized caching within an
office.

-Vivek
Received on Sun Dec 23 2001 - 08:15:09 MST