 RenCom interview about the TT server test from past Sunday [message #441065] |
Tue, 07 December 2010 15:28  |
 |
EvilWhiteDragon
Messages: 3751
Registered: October 2005
Location: The Netherlands
Karma: 0
|
General (3 Stars) |
|
|
| Goztow wrote on Thu, 02 December 2010 13:53 |
1. Can you give some statistics on the server test?
- number of players that joined
- max stable player count
- number of crashes
- number of cheaters caught (you never know  )
- etc.
|
First the requested stats:
We had about 55 players at the start, but due to some bandwidth settings issues that caused low SFPS and high use of bandwidth. After that experience we capped the server at about 40 players which was pretty much ok in regard the testing.
I'm not sure on the number of crashes, but I think the server crashed like 4 or 5 times. Another issue was that sometimes the server crashed most of the 3.4.4 clients. We are still working out what was causing this, but it was pretty annoying.
As the CPU hit 25% (quadcore, so one core was under full load) the SFPS obviously dropped at some points.
We used about 2,5 GB upstream today (the whole day) and about 150 MB downstream. Average bandwidth usage when the server had 40+ players was around 2,5 MB/s so it appears.
| Quote: |
2. Did you consider the test a success?
|
Yes, there were some serious issues, like the clients crashing and the server is not yet stable. Another thing that popped up was that the SBBO settings of the clients. With normal Renegade it is bugged so it does not matter too much what you set it too (sometimes works, sometimes fails; and lag is rather random), but with TT it works like it should and actually limits the bandwidth to that value. The side effect is that "misconfigured" clients will get flooded/DOS'ed or receive rarely any updates. For example, may players still use a SBBO setting of 56K, which means that the server will regard them as players on a 56K modem and send out very few updates. Later during the test, we corrected these values on the server side which fixed some of the lag problems, but still some lag problems remain.
| Quote: |
3. Where all server side TT-features included during the test?
|
Yes, we had the "new" features on, although it might not have shown as much as it would when everyone would use 4.0 clients. Of course, the clients are not yet available, but particularly because we want to retain backwards compatibility, this was a really useful test. After a few crashes/restarts, we enabled SSGM 4.0 as well. The stability of SSGM 4.0 seems fine, it behaved much like the earlier tests.
| Quote: |
4. How did players experience lag, compared to the normal FDS?
|
There was more lag, and higher CPU loads as well. We think the lag may be related to improper priorities being set on the network traffic. That means that if the server does not have sufficient bandwidth to send updates about every object in the game, it tends to send updates for uninteresting objects (thins at the other side of the map) rather than interesting objects (such as your own soldier). Near the end of the testing session, our coding-hero StealthEye was wandering around in-game, attempting to debug this issue.
| Quote: |
5. How many (release) blocking errors remain in the fds? Any idea when these will be solved?
|
Hard to say, we can be rather certain that at least the crashes have to be fixed before a release is possible. It is possible that there are more bugs, StealthEye and jonwil are now looking in some of the client and server crashdumps. Perhaps we will request more crashdumps for players, but if so we will post a request for it on the forums.
| Quote: |
6. Will there be more tests (for a longer period?) before release?
|
More tests seem likely now, as the lag issues need to be resolved before we can release a public server beta, and a huge test can decide whether they are indeed solved. We intend to do this as fast as possible, but again, it really depends on the actual causes of the bugs and how much time it takes to fix them.
While all this may sound rather negative, it's actually a good score if you know how much of the code has been reverse engineered and improved. One of such improvements is the actualy working SBBO. Most of the netcode has been RE'ed and because of this, there were some crashes. As it looks now there are not a lot of errors left in the server. Enough to say we need some improvements, but also enough to say that a release could be rather close. It's likely that we need to do a new test later on, this test will be held when we are fairly certain that the crashes and lag issues have been resolved (obviously  ). More on this subject later though!
To add to this, the weapon switch bug/feature has been changed to allow easy selecting of (empty) remote C4. It will still skip normal empty weapons though. The issue with the clients crashing should be fixed as well, this was due to a message that was send by the server to all clients, while only >4.0 clients understand the message. Older clients crashed, as a lot of people experienced.
To end the interview, we would really like to thank first of all the server owners who shutdown their servers so that the TT test server could get a decent amount of players. Thanks Jelly, N00bstories, Atomix and BlackIntel!
Besides the server owners we would also like to say thanks to everyone who joined IRC, TeamSpeak, and of course (especially) in-game. It was a really useful and fun event! Thanks testers! 
BlackIntel admin/founder/PR dude (not a coder)
Please visit http://www.blackintel.org/
| V, V for Vendetta |
People should not be afraid of their governments.
Governments should be afraid of their people.
|
|
|
|
|
|
|
|
|
|
|
|
|
Search
Help
Members
Register
Login
Home
