As the title asks really. If you had an online game for a few players (not a MMORPG) would the main factor for network performance be how close the players are to the server, or is that less significant than other things? I'm thinking a real-time, fairly fast-paced game not turn-based or RTS.

It's quite important, depending on the distances you're talking about. For example, if you live on the east coast of the US, you'll get better ping times to a server on the east coast than you would to one on the west coast. However, the difference may not be all that noticable, depending on your game.

However, when you look at the distance between Sydney and the US (say) then it's quite significant. Ping times between Sydney and the US are usually between 150 and 250ms which can be quite significant for fast-paced games.

It's a simple matter of physics, really. The distance between Sydney and L.A. is around 12,000km. Light travels at 300,000 km/s which means light takes around 40ms to travel from Sydney to L.A. thus, the physical minimum ping time is 80ms. Add all the servers, switches and routers in between, and 150ms is actually pretty darn close to the physical limit.

Geography is not important. Network topology is.

Here is an example -- we found that users were getting poor response in Spain, despite us having a server in Spain.

When we investigated in more detail we discovered that the users with poor response were subscribers to an ISP who used a telco whose upstream was a different tier 1 to us. The result was that our packets went to Paris (the nearest point at which those particular T1s peered) and then back to Spain to go to those users.

The result was that actually those Spanish users were better off using our Paris servers, despite them being geographically further away.



You won't be able to tell stuff like this until you actually have servers in places and can do network discovery; although in general, the closer you can place servers to the T1 peerings, the better. And, of course, the more they charge you for it.

Quote: Original post by Katie
Geography is not important. Network topology is.

This is a good point. Topology is usually fairly closely related to geography, but not always.

Case in point about topology: My home ISP is Tiscali "Broadband" (yes, am planning on migrating around Christmas time so stop sniggering at the back) and I was very interested to see traceroutes going from my connection to somewhere reasonably close by, such as the BBC in London.

Hmm, IIRC it went something like this:

Edinburgh -> Birmingham -> Glasgow -> Milton Keynes -> Warrington -> New York (yes, on occasion) -> London -> bounce around London -> BBC

and the Milton Keynes step of the journey was notorious for being a point of failure by Tiscali's own admission. That ISP was often as slow as a snail trapped in molasses and it really didn't take much research to figure out why. It reminded me of that Turkish lorry driver going from Istanbul to Gibraltar, using his sat nav and being directed through Paris and London. No joke.

Final Fantasy XI is a great example of international servers. Every server for FFXI (which was sold in retail stores in the US) are located in Japan, and both North American and Japanese players play on the same servers. The mechanics of the game placed a high value on being able to quickly attack and claim rare monsters as they spawned. Oftentimes, players knew the area that a monster was going to spawn, and knew within about 5 minutes of its spawn, so it came down to reaction time. Japanese players typically did not have much of an advantage over North American players, and they typically had around the same ping times, believe it or not. American players sometimes complained that Japanese players were out-claiming them because of the server locations, but it turns out some of the Japanese players were saying that Americans were getting all of the spawns, so I guess it is just perception.