[SciPy-User] [SciPy-Dev] numpy and the Google App Engine

[David]

On 05/27/2010 01:54 AM, Robert Kern wrote:

> Not really. It is not intended for such purposes. It is intended for
> the easy deployment and horizontal scaling of web applications. Each
> individual request is very short; it is limited to 10 seconds of CPU
> time. While numpy would be useful for scientific web applications (not
> least because it would help you keep to that 10 second limit when
> doing things like simple image processing or summary statistics or
> whatever), it is not a source of CPU cycles.

Besides what Robert said, I would also mention the datastore limitations 
given by the Google App Engine (no big blob of data, high latency, 
etc...) which make it quite hard to do something non-trivial even 
assuming numpy were available.

It is also my understanding that EC2 is pretty competitive compared to 
GAE (but of course GAE does more for you). I did not know about picloud 
until two days ago, and I have only used GAE for a couple of months at 
work, but picloud seems like a much more usable service for scientific 
computing to me.

cheers,

David