1

If I have a client-server architecture, is the up delay (from the client to the server) always the same as the down delay (from the server to the client)? On a case-by-case basis of latency measurement, of course I would expect there to be minor differences. However, when averaged over large latency samples, I would expect that the two delays are identical or very nearly so.

However, I suppose it is possible for a system to use one route between the client and server for sending data, and another route for receiving data from the server to the client.

Does this make sense? Is it really possible that a client-server based system can have, for example, a larger latency from the client to the server than from the server to the client? If so, what about visa versa?

D34DM347
  • 1,471
  • 2
  • 19
  • 32
pookie
  • 151
  • 1
  • 7
  • 1
    In addition to asymmetric routing, there could be greater congestion in one direction than the other, although this could vary and even reverse from one time to another. Plus some last-mile connections -- especially ADSL (usually telco) -- are inherently slower in one direction. – dave_thompson_085 Nov 09 '16 at 14:16

2 Answers2

0

No, we can not assume that the network latency will be the same in both directions in a client/server connection. Since the communication isn't really a "round-trip," but rather two separate one-way connections, it is perfectly reasonable that the traffic could take a different route when it moves in the opposite direction.

D34DM347
  • 1,471
  • 2
  • 19
  • 32
0

Yes, both variants are possible (C-S higher delay or S-C higher delay). It is called Asymmetric routing: a packet traverses from a source to a destination in one path and takes a different path when it returns to the source.

Overmind
  • 3,076
  • 2
  • 16
  • 25