Oh, it's quote fun time.
GimpyPaw wrote on Mar 28
th, 2016 at 10:07pm:
We have no idea what computers the VMs is are doing. "
FIFY
It appears that they tried to Virtualize the game servers... but they actually had no clue of what it entails in term of hardware management...
Virtualizing two servers into one hardware in just not getting a two times more powerful server... you need to make sure that both usages are compatible with each other and that
one usage does not starve the other of a specific hardware or computation power. Seems they didn't have any clue regarding that, and now are trying to find solutions tweaking the VMWare parameters. ( or Xen, or whatever Virtualizer they use )
Asheras wrote on Mar 28
th, 2016 at 11:34pm:
That does seem a bit like throwing stuff against the wall to see what sticks but at least they are doing something. I'd like a better description of what they are targeting. That would help in stress testing if we knew what to area to push. But I also understand the concept of confirmation bias and embedding.
You know the quality of the communication as well as me, no way they are going to give us enough details to actually be of a real use into trying to fix their problem.
It's like the traceroute... where we all traced towards a server that's only involved into authenticating us to connect to the game and not the actual real game ( virtual ) servers.
AtomicMew wrote on Mar 29
th, 2016 at 12:11am:
I know I was sort of joking before, but they are now literally using live servers Virtual Servers as test servers.
fify
pnellesen wrote on Mar 29
th, 2016 at 1:04am:
This seems to be a pretty pathetic admission that they really, truly have no idea at all what causes the "entire party frozen until killed by totally unaffected mobs" type of lag that they have laughed off for years (other than they seem to have made it worse by moving to the new servers.)
Now that I had time to ponder about it, it looks typical of a VM starvation... It's either one VM starving another, or they hardware assignment to the VMs badly done.
Basically, there's an event ( say a Guild Renown earned ) that needs some computation power and to use some hardware... but that computation power or that hardware is used by another VM, that can't free it until it has finished doing what it's doing... So the VM is starving for the resource, ( while we hang in midair ), and get the thing done once the resource becomes available.