Server processor usage?

[yurch]

What are the specs on the server boxes out there? Most of them seem to really have problems with 8+ players - extreme packetloss for everyone.

[ClaytoniousRex]

Our TBG servers for the public test are all Pentium 4's WITHOUT hyperthreading. The lack of a 2nd core can cause slowdowns when lots of bots are doing pathfinding. However, packet loss is usually not caused by the CPU bogging down but by the network connection. The best cure is to connect to a closer (lower ping) server, which should be a lot easier after the public test (at least within the CONUS).

[yurch]

Perhaps it is not direct loss then, but losslike behavior. Arpinium station seems to be the biggest offender, with just about every player complaining about hovercraft spinning out, long periods of 'server not responding' and text messages often not getting through. Especially when there are 10ish or more players and something like smoke missions firing. If it was just me, I'd think otherwise.

Under 8 players, there usually is no problem at all.

Is it possible there's something across the clients (or server) that's dumping packets that are too large? I really can't think of any other reason why it should be so consistant with high player counts.

[ClaytoniousRex]

Arpinum Station happens to also be:

1. A fairly busy http web server

2. The account server

3. The master server

4. Our primary debugging server

So it's easy to believe that this is where you get the wonkiest behavior. After the public test we don't intend to host a game server on it at all anymore.

[Rampage]

Yurch - there's a site called DSL Reports (or maybe it's Broadband Reports) - anyway, they've got a collection of utilities you can use to identify each router between you and any destination computer, and then determine whether any of these routers are dropping packets.

[weasello]

I ran several tests during the 10+ usage time when I was online with Yurch (on Arpinium Station) and all the hops seemed consistent and normal except for the end server - so I'm assuming the server couldn't handle the load at all. The in-game ping report also went from about 100 ms to 300 ms during this period.

[Type98]

Just another data point:

Every time the DT client joins a server, the client network usage will spike over 200Kbps ( last connection was around 278Kbps peak). After that it settles down to around 20-30kbps.

[pilgrim]

I mentioned this to Rune on the first day I started playing the beta test, but I think I'll bring it up again. Every FPS I've ever played has had a way to show pings of all the players in-game, and I think it's a valuable feature. If I start experiencing lag and/or packet loss, I want to know if it's (1) the server -- the pings of all the players will be high -- and then I can find a different server, or (2) me -- only my ping is high, and I know that it's just the wrong time for me to be connected.

[weasello]

One thing I really enjoy about this game is the lack of people saying "OMFG LAG!" or "PING TOO HIGH THIS SERVER SUXXX!"... I wouldn't mind seeing a ping for testing purposes but I don't want it in the final, public product.

[Rampage]

Weasel - you sound like a Battlezoner. I know the lag thing is prevalent in lots of games, but it's a real crippler in that game. Lots of vehicles moving around, some at very high speeds, lots of projectiles requiring vertical trajectory calculations. Any number of times I've been hot on an opponent's tail, shooting him up, only to have the screen update and find him nowhere in my line of sight and me with no ammo left..