Memory and CPU allocation?

[N. Rosencrantz]

Hi
I'd like to know how memory and CPU gets assigned to an application instance.
Say I deploy an application and there are no users for some hours. Then that instance will "shut down", get garba collector or likewise and then turned on if there is access again. Is this correct? So if there is access, what is the minimum CPU and memory allocated to an instance? I read somewhere that it's 128 MB and 600 MHZ is that correct and how does it compare to other memory and CPU management models?

Thank in advance for a reply /Niklas

[Amy Unruh]

hi,

Backends allow configurable cpu and memory (see http://code.google.com/appengine/docs/python/backends/overview.html), and with the upcoming 1.6.1 release, frontends will allow configuration as well (see this thread: http://groups.google.com/group/google-appengine/browse_thread/thread/8d46c99e44268b60/b4f6269552bf14f9)

--
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/-/1A-gNBPyoo0J.
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.

[N. Rosencrantz]

Thank you Amy for the reply and the links. I would like something to compare my other installations with
ie I had a virtual linux server with 256 MB RAM and 1 GHZ CPU or likewise and I concluded that my Java app needed at least 512 MB memory. PHP I could run with only 256 MB. So I wonder if there are numbers like that I can compare with app engine for instance saying that with a virtual linux server 256 MB RAM wasn't enough but in the app engine environment 128 MB is enough even for a Java application in the cloud.
Best regards \niklas

[Brandon Wirtz]

Comparing Instances of GAE to instances of EC2 is like Comparing a 5HP Briggs and Strat Engine to having 5 Actual 4 legged horses.

5 horses will haul  5 people a lot farther a lot faster than a gokart with 5HP, but no amount of Horses will accelerate an RC car to 60 Miles per hour. 

 

Same for the Memory.  Unless you have a set amount of data you know is going to be in “ram” you aren’t going to use “ram” you are going to use MemCache.

 

You aren’t going to use I/O so your hard drives aren’t going to chew CPU cycles so you can’t compare Cycle to Cycle.

 

Doing these kinds of comparisons will not help you develop good apps or plan for scale.  Right some Skeleton code to do performance testing on a free instance.  Examine how concurrency through the use of the API’s work.

 

Remember on an EC2 Instance you burn CPU cycles for everything from Network, SSL Computation, to Data processing.  On GAE if you use the API’s the CPU sits idle or processes other threads.

--

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/-/OdUu6hALckgJ.

[N. Rosencrantz]

Thank you for the information. It's very positive that we can focus on application development instead of system administration and server tuning.
I went from physical (colocation) to virtual (godaddy) to cloud (appspot) and I recoded my app from Java to python since I wanted to make an early migration to appspot at the time before Java was available for app engine. The app I migrated from virtual server needed 512 MB therefore I was surprised when I read that appspot instances get 128 MB and 600 MHZ and maybe those numbers were something else and I misunderstood.
Regards,
Niklas