Testing alternate network settings this week (11/13); requesting feedback

[qan]

A server setting that could induce additional lag was found to be outside typical parameters. This is probably a relic from a time when TW had a much larger population.

This week, we're testing with a lower value (the default one). Specifically, the setting affects how position packets are sent.

Please take note of any objective differences (latency/ping, packetloss) as well as subjective ones ("smoothness" of movement, opponent positions jumping more/less, the "hanging" effect that sometimes occurs especially when people are thrusting, etc.). Post what you find.

If there are improvements or no noticeable differences, we'll keep it. If there's tolerance for possible latency issues, we could also drill down further (during the week only so as not to affect league matches) to try and optimize performance.

[Rab]

It's good to see something being tried, thanks qan.

I will do my bit by playing and reporting what I see.

First, some thoughts on the "before" situation...

The server has had +20ms more latency (uk) (80->100) and about +0.1% S2C than usual (0.1->0.2) for the past few months.

When twdt was starting I requested for the server (windows) to get restarted in the hope that it might help, and to generally give us the best situation for competition. Staff were able to restart the game server, but they didn't have access to restart windows (understandable).

Has windows been restarted? And what's your general take on that?

If windows was restarted that might result in a false positive outcome of this change (and I just generally want to know). It would be handy to have commands to check game server uptime and windows server uptime in game to avoid having to ask.

Should we expect this setting change to affect spectators and players in the same way, or differently (in which case we must observe the change as players not as spectators)?

[qan]

Last boot was 9/24. Did you see the increase before or after that? We don't reboot the Windrows server very often because it requires coordination with Chaos, and because of an annoying "feature" that has to be handled manually.

There are some differences in how position packets are sent to spectators vs players. Spectating will still give useful subjective data, but playing is better. I'm especially interested to hear if this changes anything with jav dueling performance.

[Rab]

Last boot was 9/24. Did you see the increase before or after that?

My gut feeling is it got worse after that. I don't have data to back that up unfortunately. Maybe other ppl can chime in to agree/disagree.

[2pac]

the hub of the internet is where AWS is and that is Northern Virginia US EAST, not NJ.

This should fix the lag issues if it was put in the place where majority of the internet routes to anyway.

[Rab]

Waiting until the end of the week to post feedback, but an early request: it would be helpful to have the !lag command on all matchbots, because without it I can't tell if I'm evaluating the server or the player is just a dirty lag cheater.

[Rab]

I'm giving up trying to prove anything with this this week, the game is just too dead so there are no games to play. Generally I think it's fine to keep these settings, barely any difference, it could even be a hoax I can't be sure anything actually changed.

My notes:

My connection:
Ping to London: 6ms
Ping to New York: 74ms

All times UK (ET+5)

Mon 13 Nov 2023
- 5pm: ping: 100, ploss: 0.2 0.2
- 6pm: ping: 80, ploss: 0.1 0.1
- 7pm: ping: 70, ploss: 0.1 0.0
- 8pm: ping: 90, ploss: 0.1 0.0
- 9pm: ping: 100, ploss: 0.1 0.0
- 10pm: Spectating jav elim, looks laggy.. kama has 170ms, roiwerk has 1.2 s2c. Can't judge the server due to laggy players. Note that the !lag command was useful here, but it doesn't exist in twjd3, which makes it hard to tell what visible lag is down to the players or the server.

Tue 14 Nov 2023
- 7pm: ping: 90, ploss: 0.1 0.0
- 8pm: tried some pub, a bit frustrating, ppl aren't dying, movement isn't remotely smooth, but it's kama again so idk, no !lag command here to check the other players.
- 10pm: Spectated wb mmr, can't play because of stupid res limit. No lag cheaters here I think. I can't tell the difference in general movement. I notice that ppl aren't complaining, if the impact was clear and negative ppl would be talking about it.

Wed 15 Nov 2023
- 6pm: briefly played pub, feels fine.
- 7pm: ping: 100ms, ploss: 0.0 0.0
- 9pm: game is dead af, nothing to test

Thu 16 Nov 2023
-

General Observations
- I can't really tell the difference.
- Possibly fewer eats.
- At least it doesn't seem worse.
- Some players have unacceptable lag and this setting change won't fix that.

[qan]

Thanks for taking the time to observe, Rab.

It sounds like it should at the very least be fine to run through the weekend. Maybe the larger player count and more games will lead to more observable differences there.

We can look at further tweaking in the future, too. This initial step was mostly to get things back to what is standard on subgame2 servers. I wasn't expecting a day and night difference, but it's encouraging that it likely hasn't made it any worse.