I don't know how aware devs on this are, but there was a big important event this saturday -- Deneb Run #100

There were around 20 racers, 3 MCs and too much pirates participating, and it gave a good impression what could VO be, if playerbase were more active.

It was not excellent.

Event started when all 20 racers jumped from Odia M-14 to Odia B-13. 20 pilots is not that big number, patrol or blockade squads of even larger size will be common Soon(tm). But when 20 people hit their activate key, there were a huuuuuge lag before sector loaded. It was something around 20 seconds. And until racers split, loading times were really bad (and even after that they were worse than usual)

So, it seems like the game is not actually ready for activities of such big groups, and I thought the devs should pay their attention on what happened server-side during that event. Such stress-tests can not happen often, but their results should be valuable

Aleksey good point. I was getting major lag until we reached serco space. By that time most players weren't in the same sector jumping about the same time. At least the game didn't crash though.

Maybe we should have more runs with 20+ players to see what happens.

Well, I suppose this has something to do with pushing data down the pipes unto us, rather than hw/sw scaling problems.

I'm sure Incarnate and the others have made deep stress tests on the test server with lots of remotely controlled no-GUI fake clients stressing the system.

Because they did, didn't they?

Right?

Hello?

*chirping crickets*

Sector loading delays kind of suck now, and it all revolves around various hacks to deal with the fact that the MySQL client library blocks during queries and we need to do soft-realtime simulations in each sector, so blocking is out of the question. Right now it amounts to queries on each cluster box being serialized.

So if someone warping in has a big inventory (which should be much less of a problem now, actually) everyone has to wait until that's loaded. I don't really see any other reason why loading would be slow in that specific instance (this was about 4:45 PM CDT, right?) but I'm still looking.

There are several solutions for this on the table (primarily, stop loading stuff we don't need in every sector) but right now we're dealing with some integration problems leading to embarrassingly frequent downtimes. The whole point of what we're doing here is to vastly improve scalability, once the damn thing works.

Syylk: Actually we have done that, but not in a while.

a1k0n: thanks for the info. Jumping for me has gotten better overall. Even though i have about 3000+ widgets all over the universe.

We just thought you should not that 20 people jumping in any sector close to the same time, basically kills the jump. I guess my first few jumps i lagged 20 to 30sec. then it got better.

[ ] a1k0n
[ ] a1k0n
[ ] definitely a1k0n
[x] MySQL

The Jump was 21:04:47-21:04:50 GMT according to Himo's log: http://vendetta-online.com/x/msgboard/7/16747?page=6#215877

Heh, someone was holding down enter the whole time. But wow, that's really.. long. Hm.

Someone activated this alias at the start of the Run, and didn't use Enter during the race at all :)
alias aaction "alias act 'Activate; wait 0.001 act'; act; alias toggleaacti aactioff; echo '******** AutoActivate activated!'"
alias aactioff "alias act ''; alias toggleaacti aaction; echo '******** AutoActivate deactivated'"
aactioff
bind 4 toggleaacti

OMG. An activate DDOS!

Quick, everyone buddy invite Inc 100 times a second next time he's on!!!111!!thebigoneoneone

Actually maybe setting delay to 0.001 has no sense. My ping is around 180-200ms anyway :) But several people adviced to set it as low as possible. Does it hurt server load?

Maybe a little. I doubt it. I think only one gets sent per frame anyway but I'm not sure. But yeah, 0.2 is a little more sensible if you really want to do that.