Forums before death by AOL, social media and spammers... "We can't have nice things"
|    comp.protocols.tcp-ip    |    TCP and IP network protocols.    |    14,669 messages    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
|    Message 13,649 of 14,669    |
|    Jorgen Grahn to Hans Rosinee    |
|    Re: HTTP echo-service?    |
|    10 Oct 10 16:56:03    |
   
   From: grahn+nntp@snipabacken.se   
      
   On Sun, 2010-10-10, Hans Rosinee wrote:   
   > I'm in the need to test a client-server setup in the case of a very slow   
   > server response. A large number of thin clients chatting to a heavy loaded   
   > server via the HTTP protocol. The server could be located in another country.   
      
   I guess that last sentence means "I need to check what happens if the   
   server is far away from the clients, small bandwidth, high latency"?   
      
   > My idea is to setup a false DNS-server on the local network (where all the   
   > clients are) and thus make the clients talk to the server (on another   
   network)   
   > via a some sort of HTTP echo-service or "Man In The Middle"-service somewhere   
   > out there. It is to much work to reconfigure all the clients to talk to this   
   > echo-service. Hence the false DNS-server (ettercap?).   
      
   It is probably simpler not to touch the DNS, but use iptables or   
   whatever your gateway is running to rewrite the packets to host A   
   (your server) to host B (the "echo-service").   
      
   > The "echo-service" of MITM should be configurable on   
   > delay, bandwidth-limiting, no response etc.   
   >   
   > Basic flow:   
   >   
   > 1. Client does DNS-lookup for server-x.com   
   > 2. Fake DNS-server returns 'A' record for echo-server.com   
   > 3. Client does a "GET" + a "host: server-x.com" header to echo-server.com   
   > 4. echo-server.com passes this on to server-x.com that returns the result   
   ASAP.   
   > 5. echo-server.com withholds the result for a specified time before client   
   gets it.   
   >   
   > But I'm at loss finding if such "echo-service" exist and where that   
   "somewhere" is.   
      
   What you call "echo service" is normally called "HTTP proxy". (And   
   when people speak of "echo service" I normally think of the TCP and   
   UDP echo services, which do something completely different).   
      
   Your clients may or may not be able to use a HTTP proxy.   
      
   > Anybody got the idea what I'm talking about or is the idea totally feeble?   
      
   :-)   
      
   I think you need to clarify what you want. Some questions I asked   
   myself while reading:   
      
   - What is the system you're testing -- a client, or the whole system   
    of N clients and a server?   
      
   - Why don't you have enough control over the clients so you can   
    reconfigure them, if they're part of your test bed?   
      
   - Are you allowed to put extra load on the server,   
    stealing its CPU resources or SIGSTOP the server software momentarily?   
      
   - If you're testing the client, why isn't enough to test one at a time?   
      
   - Why don't you consider running the proxy ("echo service") yourself,   
    so you have control over its behavior?   
      
   /Jorgen   
      
   --   
    // Jorgen Grahn
|
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
(c) 1994, bbs@darkrealms.ca