0

I am attempting to implement a proxy server and a proxy client in C (separately yes!). So in this scenario, we have a client, a proxy client, a proxy server and a server.

The proxy client accepts a request for a specific host (server) and port number from a client and opens a single connection for that particular host/port pair with the proxy server.

This connection is persistent. If any more requests come for the same host/port combination, it must be sent through the same connection. The requests are numbered with id's 0,1,... in each connection to distinguish one from the other.

For every request (not connection. there may be multiple requests per connection.) received from the proxy client, the proxy server opens a connection for each request to the destination server.

However there needs to be a limit to the number of connections that can be opened to the destination server. I want to know how I can implement this limit.

Say I specify maximum number of connections that can be opened to h1 destination server to be 2. If a third connection comes in from the client proxy, what should the server proxy do?

  1. Should I implement some kind of queue to store incoming requests, if there are already 2 open connections to the destination server?

  2. If yes, then should there be separate queues for separate host/port combinations? And how would such a structure be implemented in C?

I'm feeling incredibly lost right now. Would really appreciate some guidance here. P.S. I'm coding in C using sockets.

sue
  • 1
  • 1
  • How do you plan to handle different clients? When counting connections request from the clients: Do you distinguish between connection-requests to the **same** ip/port from **different** clients? – alk May 13 '13 at 16:53
  • Why do you want to limit the **number** of forwarded connections? Limiting the **bandwidth** sounds more useful to me. By using some kind of _chunked_ encoding (where you send blocks of the type {address+length+data}), you could use a single TCP/IP stream for a big number of connections. Just an idea. In case you want to limit connections, I would simply have the relevant thread waiting on a condition variable for a free "slot". You can keep a hash table using host+port as the key. – Raúl Salinas-Monteagudo May 13 '13 at 21:55
  • Thanks so much for your replies! So i have this idea to use a thread pool. Would that be reasonable? For each host/port combination, I initiate a thread pool in which the number of worker threads=maximum number of connections. P.S. esperanto, I have to limit the number of connections. Limiting the bandwidth actually does sound reasonable but that's not the requirement. – sue May 14 '13 at 14:58
  • alk, if it is a request for the same ip/port from different clients, client1 and client2 will send their requests to the client proxy (through separate connections of course). The client proxy will then forward the requests to the server proxy using a single connection. In this case, the requests will be numbered 0 and 1, respectively. I hope that's what you were asking. – sue May 14 '13 at 15:08
  • So that's exactly what I did. I created a thread pool for each incoming connection. So if my maximum number of connections to the origin server is 2, it would create a thread pool of 2 threads. Any incoming requests would be queued and each thread would pick up a request from the queue. My proxy server is working just fine now. Cheers! ***happy now*** – sue May 17 '13 at 13:24

0 Answers0