vServer test

For things that have to do with those crazy test servers... and yeah. By request of z-man, and, of course, you gotta obey...

Moderator: Z-Man

Post Reply
User avatar
Z-Man
God & Project Admin
Posts: 11624
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

vServer test

Post by Z-Man »

I'm curious whether a virtual server (those things that are not real servers, but only running in a virtual machine on some big box shared by many vServers) has enough power to host an AA server. I got a three day trial setup for one by a local company and started a fortress clone on it. The server's name is
"vServer Test Fortress"
and runs identical settings and code to CVS Test, only without the annoying restarts. There isn't enough disk space for ongoing recordings anyway, and I could not debug them there. So BUG cries there are mostly unheard.

Throw at it what you've got and complain here if the server's performance is not adequate.

With four players, the CPU load was at 25%, all with 50 FPS simulation rate, which I'd consider the lowest value that allows quality play. That makes a fully busy server just about possible.
User avatar
Z-Man
God & Project Admin
Posts: 11624
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

The data from the last post seem to be worst case. I've played there during several times of day, and against humans, it always was extremely un-laggy. Some players called "Lag", but that was always during perfectly predictable situations and packet loss tolerance it set pretty high, so I suppose it was rather surprisingly little rubber that killed them :)

Pings are excellent (for me). From the university computer, ingame ping is indistinguishable from CVS Test, which runs on the frigging same machine as the client.
User avatar
Z-Man
God & Project Admin
Posts: 11624
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

Looks like vServers are also known as Virtual Private Servers. Anyway, the test server has been terminated yesterday.

Analysis: count the chat lines containing "lag", divide by total number of chat lines. For the vServer logs, I get
lagginess=45/5841 = 0.77%
Which is on the same level as for CVS Test, where three random recordings had a lagginess of
0.85%, 0.74% and 0.64%. (Yes, I filtered out the player named "lag"). Which probably means the real lagginess is masked by the high level of brain lag, finger lag (which I didn't filter out), residual internet lag and people just using lame excuses.

So it looks like a vServer is a viable option for an AA game server. The talk about vServers not being suited for stuff like that probably is FUD from the side of full root server providers :) For reference, not recommendation (for that, the test really wasn't long enough and no comparison was made with other providers), the test server was hosted by www.server4you.de . There is also www.server4you.com , but they have completely different offers, most notably with less included bandwidth per month.
Last edited by Z-Man on Fri May 12, 2006 10:13 am, edited 1 time in total.
User avatar
Lucifer
Project Developer
Posts: 8683
Joined: Sun Aug 15, 2004 3:32 pm
Location: Republic of Texas
Contact:

Post by Lucifer »

Well, my personal impression with 6 players was that it performed a bit better than my own server with 6 players. :) So, less lag bombs, to which my server is prone. My ping was a little higher than CVS Fortress, but that's probably just random chance more than anything else.

I looked at the american site you linked (note: can you edit the comma out of the link?). It looks like you have to go to their Pro service before you get a shell account, and they must have some of the worst marketing flunkies writing their website copy. Absolutely nothing that I wanted to know was anywhere to be seen, and I'll be damned if I waste time talking to a sales rep when I'm just window shopping. Anyway, they seemed geared towards web hosting and the plans offered all look like regular web hosting plans, only high dollar hosting plans. And their try before you buy was a joke. 1 hour of try before you buy? What's that going to tell me? That I can see their server? Yippee. I wasn't sold, but I'd be happy to write them a website that's worth more than a ... (family forums).
Image

Be the devil's own, Lucifer's my name.
- Iron Maiden
gnorty
Core Dumper
Posts: 187
Joined: Wed Nov 02, 2005 2:45 am

Post by gnorty »

How much improvement in lag by running the client and server on seperate machines on a LAN? ie if I ran a server on lame.gnorty.net and a client on leet.gnorty.net would there be a significantly better connection from the internet?

I am guessing that leeching CPU from the server to feed a hungry client is a likely cause of lag.

On a slightly different subject, how much bandwidth does CVS fortress consume?
User avatar
Lucifer
Project Developer
Posts: 8683
Joined: Sun Aug 15, 2004 3:32 pm
Location: Republic of Texas
Contact:

Post by Lucifer »

Yes, leeching cpu time from the client can really hurt the server. Back when I played on the server, 8 players was too many. My server could only reasonably handle 5-6 while I was playing, but players reported it fine when I wsn't there for more players. Nowadays it's different, I play on a different machine than the server, and it handles predictably all the time.

Also, I'm talking about that 800mhz Duron machine, slower than molasses when you get right down to it. My fps would drop with each new player on the server, and a full server brought my client to 22fps with all the crap turned off. Kinda frustrating.

Ironically, the problem smoothed a bit if I just played other servers. There was still cpu time sucking going on while the two processes fought for cpu time, but it wasn't so bad. So I could play Swampland reasonably while my server was full. That was, if anything, even more frustrating, because if my server was full, I wanted to be there to enjoy it, and I couldn't be there without destroying it. :(
Image

Be the devil's own, Lucifer's my name.
- Iron Maiden
User avatar
Z-Man
God & Project Admin
Posts: 11624
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

gnorty wrote:On a slightly different subject, how much bandwidth does CVS fortress consume?
Peaks as given by the statistics output are around 30kbyte/s out, 10kbyte/s in. Real peaks are about twice of that, and I hope the daily average is only a third :)

I hadn't really looked at the US site :) Beyond the "oh, the'yre cheaper. Oh, you're getting only 1/10th of the bandwidth." point. All the "not available" crosses in the offer details aren't really encouraging, that's right. Where was the "try before you buy" option? I couldn't find it. It's probably some completely different company that has nothing to do with the German server4you (although the page layout and logo is the same).
Post Reply