home *** CD-ROM | disk | FTP | other *** search
- > Network DOOM Traffic Analysis
- ...
- > While it is true that the first versions (1.0 and 1.1) of DOOM used
- > "broadcast" packets to communicate with other DOOM workstations, that
- > method was changed in version 1.2 to use "directed" packets. Directed
- > packets have less impact on LAN traffic than broadcast packets.
-
- On something like ethernet it shouldn't, since a broadcast simply means
- that all the connected machines can get the data from one transmission.
-
- If the LAN port works as I think it does, a BAD MOOD network using that
- should probably use broadcasting.
- MIDI networking on the other hand will need a ring based configuration.
-
- > The below table illustrates exactly how much network traffic is generated
- > by DOOM in a 2, 3 and 4 player game.
- > __________________________________________
- > No. of Players | 2 | 3 | 4 |
- > ------------------------------------------
-
- > Packets per Sec. | 35 | 70 | 105 |
- > (per PC) | | | |
-
- That's the 35Hz world update frequency I've mentioned before.
- I don't believe the ideas about updating once per displayed frame on the
- slowest machine are going to work well.
-
- > Packet Size | 72 | 72 | 72 |
- > (bytes) | | | |
-
- That's a lot more than I would've thought necessary.
- Perhaps the network set the size.
-
- > interesting to note that there is three times more traffic generated
- > going from a 2 player to a 3 player game, and six times the traffic
- > going from a 2 player to a 4 player game.
-
- This is of course obvious when you're using point to point communication.
- With two machines, each will only have to send to one other: 2 messages
- With three, there's two to send to: 3*2=6 msgs
- With four, three 4*3=12 msgs
-
- > Even though DOOM uses little bandwidth, it does generate a lot of packets
- > in a 3 and 4 player game. Because of this, network DOOM players should
-
- This will not be so in our networking, since point to point communication
- is totally unnecessary in a dedicated network.
-
- The major problem for us, assuming we go the MIDI way, will be latency.
- A 72 byte package sent over a MIDI channel takes about 20ms, which over a
- 5 machine link sums up to 0.1s.
- If all the links are used all the time and a machine in the ring sends out
- a new report as soon as it gets its own last message back, it would still
- take on the order of one frame update to get the positions of all players.
-
- Hopefully we can reduce the size of the packets considerably.
- At a dozen bytes or so, things would probably work rather well.
-
- The LAN (if it handles broadcast) would be preferable since the bps rate
- is likely to be a lot higher than the MIDI can handle. The normal serial
- port is not bad, but without special cables it can't handle more than
- two machines.
-
- --
- Chalmers University | Why are these | e-mail: rand@cd.chalmers.se
- of Technology | .signatures | johan@rand.thn.htu.se
- | so hard to do | WWW/ftp: rand.thn.htu.se
- Gothenburg, Sweden | well? | (MGIFv5, QLem, BAD MOOD)
-
-