Re: An expected latency formula (fwd)

On Mon, 20 Jul 1998, Yee Man Chan wrote:

> I am trying to derive a formula to find the expected latency of a client
> using Squid.

Note that Squid pipelines replies. That is, the total client response time
may be (and in most cases is) less that the sum of the components. Connection
setup time is also very important (and may be different for different types
of requests/replies). With persistent connections in place, the whole thing
becomes even more complex. (Not to mention cooperative caching.)

Also, to validate your formulas, you would need logs that are more detailed
than standard Squid format. [ You can look at my profiling stats, but they
are based on caches that do participate in the hierarchies. ]

I bet that even if the magic formula exists, the actual coefficients will
vary a lot from cache to cache and from time to time, making "average"
expected response less useful.

Finally, browsers often submit several (4) requests at once. Thus, the user
perceived response time has very fuzzy connection with the actual response
time of individual requests (not to mention other human factors involved).

Alex.
 
> The formula is derived based on the following assumpitons:
>
> 1) Simple client-proxy-server model. No co-operative caching.
> 2) Disregard network failure and hardware failure.
> 3) HTTP headers are identical in size regardless the extra options
> specified
> 4) Memory access time is neligible when compared to other I/O time.
> 5) HTTP reply other tahn 200 or 304 is disregarded
Received on Tue Jul 29 2003 - 13:15:51 MDT