[google-appengine] Re: App Engine getting slower and slower
After experimenting with various F1-F4 frontends, I meanwhile have migrated the workload of my app to public facing B1 backends, so it can be accessed remotely. As you can see from the attached screenshot the latency has improved greatly compared to using frontends. I now have again approx 150ms latency compared to up to 1 second during the last days using frontends. Not sure if this is because backends operate in a new / not so busy cluster or if simply the scheduler for backend traffic is handling it with less delay. It's obvious that there must be a difference. -- 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/-/JeKVmeLk_P4J. 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. attachment: backend_frontends.jpg
[google-appengine] Re: App Engine getting slower and slower
The scheduling is completely different for backends, and if you're using only one instance there's simply no autoscale scheduling. Requests are simply put into a waiting queue 'till the backend is free (or they time out). -- alex On Thursday, July 12, 2012 2:29:52 PM UTC+2, Marcel Manz wrote: After experimenting with various F1-F4 frontends, I meanwhile have migrated the workload of my app to public facing B1 backends, so it can be accessed remotely. As you can see from the attached screenshot the latency has improved greatly compared to using frontends. I now have again approx 150ms latency compared to up to 1 second during the last days using frontends. Not sure if this is because backends operate in a new / not so busy cluster or if simply the scheduler for backend traffic is handling it with less delay. It's obvious that there must be a difference. -- 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/-/utc2-QSe4QYJ. 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.
[google-appengine] Re: App Engine getting slower and slower
I'm using more than 1 backend for redundancy reasons. I was also experimenting in addressing a specific backend directly, but couldn't notice any difference in latency compared to addressing the whole backend pool. By using backends I now see log entries in the range of 20-30 ms for processing a request that is enqueued to the taskqueue compared up to several 100 milliseconds that were consumed (or kept in queue) on frontends. Running the backend option I can now serve the load that required much more frontend instances (or at least that's what the scheduler thought would be required). The only downside is that there will be no unlimited autoscaling in place as with frontends. Looking forward for GAE to fix the frontend issues, as latencies like these don't do the platform any good: http://code.google.com/status/appengine/detail/serving-java/2012/07/11#ae-trust-detail-helloworld-get-java-latency -- 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/-/A5hbiZiO0lMJ. 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.
[google-appengine] Re: App Engine getting slower and slower
By using backends I now see log entries in the range of 20-30 ms for processing a request that is enqueued to the taskqueue compared up to several 100 milliseconds that were consumed (or kept in queue) on frontends. Impressive, thanks very much for posting your results. Did you happen to verify the improved latency is seen externally to app engine, too? I ask because I know the log entry ms value includes time waiting in the dispatch queue when frontends are used, but I haven't verified whether it's included when backends are used. -- 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/-/6TyxVeKKorcJ. 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.
[google-appengine] Re: App Engine getting slower and slower
In my case, Appengine gets faster overtime, so I start hitting idiot minutely or hourly urlfetch limits, so from time to time I have to limit task queues by reducing their number of max_concurrent_request's -- 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/-/NYte0emA2MYJ. 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.
[google-appengine] Re: App Engine getting slower and slower
What I've observed is that a specific app may tend to get slower over time as the cluster it runs on gets busier. Eventually an app may get moved to a new cluster. This may include a few minutes of hard downtime, even for HRD apps. Afterwards performance can be dramatically better. Someone at google may be able to manually move your app if you complain loudly enough, although there's no guarantee the issue won't recur. When it gets really bad all your instances can be totally idle but many requests still block in the scheduler for hundreds of ms (or more) like you seem to be seeing. Adjusting the latency settings doesn't help because the instances aren't the bottleneck. Quite possibly my least favorite part of app engine, although on the whole I still like it a lot. -- 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/-/qi2Y7GR4xfUJ. 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.