Was thinking the Same where they've worked from nothing and created it all they may not encounter any of these but it's alway's good to stress test regardless, even Marble seen Crash's early game where so many was on.
It's really hard to stress test, you'd have to Build some kind of bots that actually ran around doing something or actually find load of people willing to log and do stuff, logging on 200 bots in SZ isn't stress testing.
Having 200 people on still does start to highlight issues as we did this on TPR test (this is when the spawn errors started flagging up) as it still takes resources to process them being there (obviously not as strenuous as 200 people pvping).
To be fair they are a lot more advanced coders then me so I dont want to come across patronising to their level of skill, just things that came across with TPR.
Do you like have a download source that isn't through your patcher? Your patcher speeds are dreadful lol. Not even breaching 1MB/s...feel like it's 1999 again
EDIT: Yeh...went down to 500KB/s...Maybe when I have a spare 10 hours i'll consider jumping on lol.
So one of the big ones, if you haven't done anything with how Crystal utilises spawns, you WILL get server side lag, when your UC goes past a certain threshold (your mobs will just stand there and do sweet f/a).
Changing your spawn system to be more efficient will sort it out.
Logging is another huge one (that people are aware of) using the default logging will cause it to appear to crash (when its just stupidly large amount of lag)
> Utilising other logging methods cures this ; TPR we log near enough every possible player transaction so we have full tractability of everything, if we ever need it.
Because you've done it from the ground up, I highly doubt you will encounter these errors either; but it tends to be 2 of the major ones from the default source!
We have already put into place spawn masks, so the server wont try and brute force random valid cells to spawn in, it will simply pick one randomly from a list, so no CPU cycles wasted there!
Saving/Logging has been fully rebuilt and is fully asyncronous, it will not holdup any of the threads processing players/packets/monsters etc.
The architecture we are running on is effectively spreading the load across several cores and we have implemented some good measures to reduce processing required
We've changed alot in the back end, hence the extended beta - we have more to test than usual
Do you like have a download source that isn't through your patcher? Your patcher speeds are dreadful lol. Not even breaching 1MB/s...feel like it's 1999 again
EDIT: Yeh...went down to 500KB/s...Maybe when I have a spare 10 hours i'll consider jumping on lol.
I have seen this happen.. i think its an issue with the patcher - if you close it and reopen the speed should go right up. When ever I've patches its been mostly at 10MB/s. We will likely use a better host come live launch (VPS we use only has a 100mbit pipe atm )
Having 200 people on still does start to highlight issues as we did this on TPR test (this is when the spawn errors started flagging up) as it still takes resources to process them being there (obviously not as strenuous as 200 people pvping).
To be fair they are a lot more advanced coders then me so I dont want to come across patronising to their level of skill, just things that came across with TPR.
We have emulated 500+ connections using basic packet bots, which the sever coped with pretty well... however we are likely to re run this test, may give them a little bit more logic to fight instead of just walk around
It's really hard to stress test, you'd have to Build some kind of bots that actually ran around doing something or actually find load of people willing to log and do stuff, logging on 200 bots in SZ isn't stress testing.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
