The time period you are talking about is minimal. Hard to tell if you "broke something" if your instances are being slowed by something, or if you just have a difference in traffic numbers.
This is my 30 day numbers for one of my larger apps. For me what happens is Datastore, or Memcache gets buggy, and that changes the score (this is visible via the errors per second) The "Instance" size won't fix slow reads from Datastore or Memcache. From: google-appengine@googlegroups.com [mailto:google-appengine@googlegroups.com] On Behalf Of Marcel Manz Sent: Tuesday, July 10, 2012 12:41 AM To: google-appengine@googlegroups.com Subject: [google-appengine] App Engine getting slower and slower The more and more applications are hosted on App Engine seems to have a negative impact on performance. A version of one of my apps is handling the same type of requests over and over and is more or less loaded with a steady request rate from a remote system. This version does nothing more than to take a request and send it off to the task queue. As it can be seen from the attached chart, the milliseconds per request seem to be increasing over time. I have tried to compensate this by upgrading F1 instances to F4, but that doesn't show any improvement as its load is API bound. The spike in the middle of the chart is a result of this incident: http://code.google.com/status/appengine/detail/taskqueue/2012/07/03#ae-trust -detail-taskqueue-add-many-latency Not only do the APIs seem to become slower (in my case taskqueue.BulkAdd), but also does the time a request is waiting for an instance to serve increase lately. For example taskqueue: There have been times where it was possible to enqueue tasks in approx 10 milliseconds: January 1st 2011: http://code.google.com/status/appengine/detail/taskqueue/2011/01/01#ae-trust -detail-taskqueue-add-many-latency January 1st 2012: http://code.google.com/status/appengine/detail/taskqueue/2012/01/01#ae-trust -detail-taskqueue-add-many-latency Nowadays one can be lucky if this succeeds below 100 milliseconds: July 9th 2012: http://code.google.com/status/appengine/detail/taskqueue/2012/07/09#ae-trust -detail-taskqueue-add-many-latency Further I see in the dashboard logs very frequent requests with an overall time of several hundred milliseconds, many with over one second, as well as requests that are even taking several seconds. These are non-loading / non-warmup requests and according appstats they promptly execute. I must assume the difference between the ms value in the logs and reported by appstats is the time the request was waiting to be served. If so, I don't understand why this is taking so long to be served. My pending latency settings are Automatic / Automatic and I always keep 1 or 2 idle instances, which are sufficient for the load the app is handling. In one of the Google IO 2012 talks gets mentioned that 1 millisecond +/- delay results in millions of dollars +/- for Google Search/Adwords. It's good that you know that, but please speed up the App Engine platform so we can profit from the same. Thanks Marcel -- You received this message because you are subscribed to the Google Groups "Google App Engine" group. To view this discussion on the web visit https://groups.google.com/d/msg/google-appengine/-/YOi0uXp2BBQJ. 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. -- 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.
<<image001.png>>
<<image002.png>>