I think if I let my costs rise by lowering the "max pending latency",
kicking off new instances, that might cause a more gradual cost increase,
than letting costs rise by increasing min idle instances.  Based on the
graph below, if a I set the min idle instances to 2, I'll have 2 instances
running 24 hours, costing $1.92 a day, per instance (not considering free
quota or pre-paid instance hours for the moment).  But decreasing max
pending latency might cost higher, but more proportional to heavy loads,
than just a flat $1.92 extra every day.  Comments?

[image: Requests/Second (24 hrs)]

On Thu, Dec 1, 2011 at 5:56 PM, Rishi Arora <[email protected]>wrote:

> I have an Android app that sends HTTP GET and POST requests to my app on
> GAE.  Why would this client occasionally get errors like "Connection to
> https://shiprack-test1.appspot.com refused".  Looks like a TCP level
> connection refusal, and not an HTTP-level error.  My app has a max idle
> instances set to 1, and I have seen that the number of active instances is
> usually always 5 or 6 during the day, peaking at ~15 every now and then.
>  At night, the number of active instances is 1, and occasionally 2.  Is the
> connection refusal a sign of extremely high latencies due to excessive
> queueing?  My min-pending latency is 10ms, and max is 2 seconds.  Sounds
> like if I reduce the max pending latency to a smaller value, I'll cause
> more instances to get spawned in response to heavy load, but I might
> address my connection refused problem?  Am I thinking along the right track
> here?
>
> Thanks in advance.
> Rishi.
>
>

-- 
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