> So the questions remains open: > - How many concurrent users do you have to support ? I'm afraid, I still do not understand what exactly are you asking, taking into account that you wrote: > I don't want answer related to processors. Then for what part of the application is the "concurrency" assumed? For internal application servers which produce custom data, or for socket layer on the hypotetical "http-gateway"? If all connections have keep-alive option enabled, we could anticipate 6000*N open connections on the gateway in every moment, where N is a number of backend web-servers. We plan to start with N=3-5, but generally speaking, N has no upper limit, and may grow several times in the future. The question of scalability itself assumes that we want the application to work under _any_ ever growing load, by simply adding new backend servers. If this wouldn't turn out to be the case, then this is not scalability, imho.
> I can add this question: > - What is the maximum allowed latency time ? 1 second. > This being said, 10 users per server is normally very small number unless the > user > are asking a lot of data or number crunching processing. Yes. The application involves on-the-fly analytical processing. > > If session timeout is set to 10 min, > > The question is: why would a session be 10 minutes ? Why not 10 seconds or 10 > hours ? This number is deduced as an avarage value of use cases and ergonomics of the application. The 10 min is given, and there is no place for "why". Of course, the settings may allow a user to change the value in a reasonable range, say 1 - 15 minutes. > You have to tell more about your application if you want good advice. This is a kind of world-wide context searching service. In future it must process 2 000 000 requests aday. > > Max. 50Kb/sec for 1 request > Then how long is a request? Normally, ~ 1 second. This includes not only the time of transfer for this 50Kb, but also processing time. But external factors may affect performans, so request may be processed in several seconds. Most of the time the data is not transmitted from our server to the browser. > 500 kilo-bit-per-second is not big. Any personal computer is able to handle > this. Of course. The problem is that this is not a pure web-server. We have internal application servers which make it impossible to pack all the stuff into an ordinary PC. I understand that we could possibly configure a single web-server to deliver incoming requests to different application servers, but this is another issue to be investigated. I don't know yet which approach is easier to implement. Best wishes, Stanislav Korotky, Russia, Moscow, GMT +3.00 -- To unsubscribe or change your settings for TWSocket mailing list please goto http://www.elists.org/mailman/listinfo/twsocket Visit our website at http://www.overbyte.be