Wow, thanks for the attention, Brett! We just did a big site update for our app, too, and have been running around trying to fix critical bugs and get users back online, too, so you've my sympathies.
I haven't been doing testing since you guys got the latency levels down, since I did a big blitz right before overhauling the app to be immune to increased latency (from the user's point of view). We still have a lot of handlers running a few hundred ms cpu, but they haven't been getting slammed like they were before, and I got a better version of my one library that was taking 1300ms to import. I think you've explained the unresolved mystery of why the handlers were getting slammed so hard. We were using app caching, weren't anywhere near 30 instances, and were getting hit bad even with no datastore access and no instance startup costs-- --but the bit about scaling instances to match sustained throughput is enlightening. I had thought that when Marzia said that prioritization was per-request and not per-handler or per-app, that meant this sort of thing (fast requests being deprioritized) wouldn't happen, and that's why I was confused. The smaller requests slowing down because of the big requests (which were mostly happening on instance startup and were getting deprioritized hard) sounds like my issue. Requests were very variable on the app, and so a sustained throughput calculation would be significantly off from second to second or minute to minute, which may not have helped things. One thing we found out we could do at the developer chat was store a lot more than we'd counted on in the memcache. Since we couldn't find any limits posted, we were being (extremely, as it turns out) conservative with memcache space. We'll be able to increase performance in many places by putting more lookup tables in there. I know that there are few guarantees when it comes to how much will fit in there, but if Marzia has said that around 100MB is the right ballpark, then perhaps some info can be put online so others can have some estimate of its capacity. After the dust's settled, I'm quite happy with App Engine performance, and the ease of use has been a dream. Thanks for all your hard work! --Nick --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Google App Engine" group. To post to this group, send email to google-appengine@googlegroups.com To unsubscribe from this group, send email to google-appengine+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en -~----------~----~----~----~------~----~------~--~---