Do F1 F2 F4 Google App Engine frontend instances really cost more

2,595 views
Skip to first unread message

Marc Hacker

unread,
Jul 17, 2012, 10:05:18 AM7/17/12
to google-a...@googlegroups.com
If F2 costs double as much as F1 per CPU hour but takes half the time to complete tasks shouldn't the total cost be about the same?

Thanks

Barry Hunter

unread,
Jul 17, 2012, 10:29:41 AM7/17/12
to google-a...@googlegroups.com
If you use instances to full capacity - yes. In an ideal enviroment

Using 2 F1s at 100% CPU is the same as 1 F2 at 100%


But a real application will rarely be completely CPU bound. Using
external APIs wont be much different, or even 'program startup' wont
be exactly double the speed (disk access will be similer for example)

On the other hand, the extra memory headroom, will allow F2 to run
somethings much quicker. (eg an algorithm could perhaps do a sort by
coping data, rather than sorting in place which would be required on
F1),

So the relationship between the two isnt linear. A F2 isnt exactly
twice that of F1. Everything wont be exactly 2x faster.

Tasks wont just execute twice as fast.
> --
> 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/-/jnf-xzkuwD8J.
> To post to this group, send email to google-a...@googlegroups.com.
> To unsubscribe from this group, send email to
> google-appengi...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/google-appengine?hl=en.

Marc Hacker

unread,
Jul 18, 2012, 3:35:12 AM7/18/12
to google-a...@googlegroups.com
Thanks Barry.

So I think you're saying that the instance hours are charged even when the CPU is not busy and F2 costs twice as much as F1 even when the server is idle between requests or when waiting for database, right?

Strangely we are experimenting with switching between F1 / F2 / F4 and seeing virtually no difference in performance of a server benchmark (pure CPU - a big for loop).  Could it be that the F1/F2/F4 switch does not take effect immediately?  Or could it be that when the server is not loaded F1 will perform the same as F4?

Thanks

Francois Masurel

unread,
Jul 19, 2012, 3:50:22 AM7/19/12
to google-a...@googlegroups.com
Have you tried to shutdown your currently running instances?  It should help.

Marc Hacker

unread,
Jul 19, 2012, 3:59:03 AM7/19/12
to google-a...@googlegroups.com
Thanks yes we found that stopping instances is necessary to activate changes in the F1/F2/F4 settings.

But we are still getting F4 at most 2-3x faster than F1 even for a pure CPU benchmark (for loop)

Douglas Pollock

unread,
Jul 24, 2012, 1:25:04 PM7/24/12
to google-a...@googlegroups.com
Yes, as it stands right now, you need to restart the instances.  You can do this by manually restarting the instances, or by deploying a new version of the app.



--
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/-/Heey-25_BqkJ.

Michael Sander

unread,
Jan 23, 2013, 12:42:44 AM1/23/13
to google-a...@googlegroups.com
I've run some tests and F2 is certainly faster than F1, especially at CPU intensive things such as generating templates. Unfortunately, it also seems to be close to double the price.
Reply all
Reply to author
Forward
0 new messages