> These numbers tell you how much a request cost, but don't tell you how > much another identical request will cost. >
The variance between my "identical" requests is less than 10%. In relation to multi-threading, I only see a variance on the downside of a Spike, when I have Idle instances. That seems to be less than 10% as well. If you have lots of idle threads raising your price you aren't optimizing your Defer Queue and steady state tasks effectively. We have seen AWS days where cluster congestion has raised our bills by 3-4x using a Specified QoS based on "latency"/time per request/request queue depth. (same as tuning the scheduler). As always the people who are complaining don't seem to understand how the Scheduler works (which is quite likely a fault in Google's Documentation/lack there of). You either pay for QoS that is X, or you say your users can have a crappy experience, and Latency won't make a lick of difference in your App. My biggest charge days are not days that "latency" spikes but days that MemCache gets slow, or seems to "dump" more often than expected. This hasn't been enough of an issue for us to implement our own Memcacher as a Backend primarily because we are using 10+ gigs of memcache and can't build an instance to do that. -- You received this message because you are subscribed to the Google Groups "Google App Engine" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.
