ons 2009-10-28 klockan 08:36 -0400 skrev Brian J. Murrell: > It would be nice to have SNMP counters that tracked cache hits and > misses in terms of the number of bytes. This would allow me to see > effective my proxy was at avoiding network traffic. There is, kind of. SavedBandwidth = cacheHttpOutKb - cacheServerInKb but remember to do the calculation on delta values (i.e. difference over 5 minutes) as both are counters and will wrap around to 0 at some point... > That said, I'm unsure how I would account for requests to a web servers > to test for object freshness, given that that is not a client request > per se but rather a cache server "management" (or overhead). The above is pretty neutral on the subject, giving the difference in bandwidth usage. the distinction between hit/mis is quite blurred which hits also using upstream bandwith (cache revalidations), and will be more blurred over time when we add support for partial responses.. Regards Henrik
