best way to build time series server using python

I'm interested to build a fast server that serves queries on time series. For example, say I have 1000 time series identified 开发者_开发百科by category name x. The server will take a query submitted by a client process and immediately return the last value associated with a particular timestamp.

For example on the client script, someone would call the following function from a time series library:

def get_ts_value( id, timestamp )
     // send message (something like message queue, remote procedure call?)
     // send blocks until value comes back
     return request_ts_value( id, timestamp )

On the server side, the process will keep most if not all of the time series file open or at least preload everything into memory so queries can be fast. Along with some sort of index information loaded into memory for quick searches.

My question is, which ones are the easiest library/modules to build the above or if there are any existing free frameworks available?

Thanks!

As far as communication goes, do you have a protocol in mind? HTTP? raw TCP?

I would personally recommend an HTTP server using http://docs.python.org/library/wsgiref.html , although there's a chance that even that is not fast enough.

You could also use an SQL server.