[Stackless] Google Application Engine Thread on Stackless

[Andrew Francis]

Hi Rana:

--- On Sat, 3/28/09, Rana Biswas <xdevice at gmail.com> wrote:

> Well I was referring to the example of WS-BPEL
> where time span is is so huge (in days or weeks), I guess
> storing the state in a DB would be a better option
> IMHO.

I think the persistent store part is the low hanging fruit. A file or a database don't make much of a difference. 

> And yes, as others have pointed out these needs
> to tested properly to figure out time taken for
> serialization and de-serialization and compare it with  DB
> read and write, specially when load is high and data size is
> large.

I missed those statements but that would make sense. However given the potential large variance in wait time, de-serialization/db write times would be tiny.

I think one problem would be thrashing as per an OS. However I think the more interesting problem from a computer science/operations management/software engineering perspective deals with persisting a programme that has a serious bug that you have since discovered:

It has just been discovered that Programme X has a serious bug. However N instances that have been swapped out. What do I do? Do I terminate them? Do I allow them to run and fail? How long does it take to the drain the pipe of bad programmes? Do I use introspection and pickling knowledge to repair the images? Is the probability of this high enough to make it a major concern? What software steps can I use to protect myself?

Cheers,
Andrew