> Hi > I'm working on a project where I need to keep response times low, but > where > the dataset continues to grow and we get thousands of requests per second. > Each request I need to process this data and respond quickly. Currently, > the data isn't large enough to slow down the response time very much, but > I > foresee this becoming a problem as time goes on. We've been using nginx > and uwsgi thus far, and it has been working pretty well, but I need some > way to process the data quicker. Unfortunately, none of the computations > can really be done ahead of time, so I was thinking of a way to break the > data down into sections and processing it in parallel. Fortunately, > though, the dataset is formed well for this, and compiling it after each > part is finished wouldn't be difficult. > > After reading about the docs about how uwsgi works with mongrel2, I had > the > idea that I could make several uwsgi "worker" servers and put a mongrel2 > server in front of them, which communicates what should be processed one > each (with zeromq), then puts the pieces back together when it gets the > response from uwsgi servers and finally returns the compiled response to > the client. > > I'm not sure if this is the best idea or if I am viewing this right > conceptually, but any feedback would be appreciated, whether it be telling > me I'm going down the wrong direction and which would be a bette option, > or > giving me more specifics and guidance on this option. > > Thanks for the patience, I know its a long one. > >
If i understand correctly, you want to split a single request and send each 'chunk' to a different uWSGI worker. The frontend server will receive back the response from each worker and will reassemble it, sending finally to the client. Am i right ? -- Roberto De Ioris http://unbit.it _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
