> And at the end of the day, if nothing else works for you, there's also
Compute Engine. Have you tried that?

Yeah, the latency between AppEngine and Compute Engine is often too high for
some of the stuff we want, and without access to the shared datastore we
lose too much along the way.



>> And there are weeks when backends will work as advertised. Then they
won't.
>
> You'd have to back that up with more than just words and links to 
> SO/issue tracker with missing config params or "more info needed"
> label :)

Have forwarded in the past. Several apps used to run almost exclusively as
backends. Used the backend for NLTK which by default doesn't fit in an F4.
So it ran in an f8 backend. The front end would proxy requests so that we
could have a pretty URL, and that would then hit the backend.  One day
magically it stopped auto-scaling. So we moved to static backends. Due to
cost we stripped NLTK down to fit in an F4. Now we don't use NLTK.

Backends are still useful for load testing, but they 503 a lot. Even when
they are working, because you don't have the control you do of a front end,
if you suddenly need 100 requests it fails.

The queue per backend is 10, and only one new instance at a time will spin
up so you only get a queue of 20 durring the time that you spin from 1 -2
and if your spin up time is 1.5s then you can only handle 16-ish request in
the first second. Not real useful if you want to process 100 articles that
will all take 90s to run.

There is not a good way to use "defer" so you can't do retries either, 



-- 
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.

Reply via email to