Gunga wrote on Mar 13^th, 2013 at 10:24am:


Gunga...you are why I keep coming back.

Quote:


True. I read Stainer's post and all I heard was this: http://youtu.be/5DmYLrxR0Y8

stainer wrote on Mar 13^th, 2013 at 1:38pm:


You're hot.

stainer wrote on Mar 13^th, 2013 at 4:32pm:

I want a Dickies ride, Big Sexy!

Dark_Helmet wrote on Mar 21^st, 2013 at 11:24pm:


You're dumb.

Flav wrote on Mar 12^th, 2013 at 4:20pm:


Based on my experience, you are overkilling a test replicating environment.  I GUARANTEE that is not what is being run for the developers at Turbine either.  Their environment is MUCH MUCH smaller.


A working game server would not need that.   If you wanted to include that then you should hire the 20,000 real humans that do the human things to stress it.  That ain't going to happen.

The Billing/accounting/authentication can be completely disconnected and forced.  That is a completely separate part of the system.

The only thing you need are three servers and a wired network with two humans running how ever many machines they think they can handle. The hardest thing to simulate is the ACTUAL load placed on the system by 120,000 internet machines from around the world and the dropping of packets and the bad handshakes.

I do agree that Turbine's handling(or lack thereof) of known bugs and then compounding them is beyond deplorable.

I forgot to come back to that thread for a while, bunch of answers:

stainer wrote on Mar 12^th, 2013 at 4:30pm:


yes, and it helps reduce the cost.

Asheras wrote on Mar 12^th, 2013 at 4:40pm:


Just the Symetrix in the 2 million range ( and the relevant SAN )... And that's one of the of the thing that cannot be really mimicked by something else sadly... you will still need a SAN.
( ok, there's cheaper SANs, but if we talk performance, you will have to have a full copy of the live system )

Terebinthia wrote on Mar 12^th, 2013 at 6:26pm:


Seriously  : no, but if you want to test thoroughly and have a test system that looks like live fairly way, there's no way around you have to pay the bill.

In my Industry ( Telco ), not long ago operators paid for a full fledged Public Exchange with 64K subscribers ( virtual )... Just because there were things that couldn't be reproduced any other way, and performance issues are different in a 16K system and a 64K system.

notajedi wrote on Mar 24^th, 2013 at 6:11pm:


I talked about a live system when I described that...
But if you want a test server you will just need at least an Oracle Server ( with the cluster layer, even if it's only a single node ), a Clariion ( to replace the Symetrix ), The SAN ( way smaller, can probably be done straight from the Clariion interfaces, thus getting rid of the SAN Fiber Switches ), The back end network and the front end network can be the same, Game Server. Billing/Accounting can be simulated, as it doesn't matter.
Still that's two servers, 1 SAN Storage Array, 1 Network infrastructure... And with that you will definitely not be able to test any performance issue as it is not representative of the live system.
( on the other hand you'll be able to test all rest. )

The bitch in QA Test is the Performance tests... as most of the time they can only be done on the live system,
In Telco we do that on special test systems, when we can, ( my employer has a few like that, we used to have one in France ) that are basically a complete copy of a live site that are here only for the performance test purpose ( well there's a few other test cases that require these systems, but they are not relevant to the discussion )

I made a few posts citing it as a router with a problem on the official forum.
( most obvious one : http://forums.ddo.com/showpost.php?p=4904257&postcount=129 )

The problem is that it's not on Turbine's side and if PNAP tells them everything works fine on it they can't do shit about it.

And as we have people like MajMal that reads here (semi) regularly I expect they got the complete explanation too. The problem is that I can't really post the full detail on the official forums ( like the WAGs I made above about the hardware structure ) as they could be construed as trying to reverse engineer the game and would get me a game ban. ( and I'm wary about sending them in PM as it might end up with the same

In the case of Today's lag spike, it can be any router between the European end of the TAT Fiber and the dodgy router we know off. Unless you know of Americans ( or Aussies ) that had the spikes too. ( and it can also be the G-Land Hamster that wanted to go on strike) 

AtomicMew wrote on Mar 24^th, 2013 at 7:47pm:



Because of this coincidence, I've long had a suspicion where this occurs is a "seam" in the map... a place where one or more developed parts of a map were stiched together.

AtomicMew wrote on Mar 24^th, 2013 at 7:47pm:


Yup, it's not the entire issue... There are spots like that all over the game and honestly I have no clue why these particular spots will generate a lag spike when you reach them.
Shroud Part 2  when running towards the chests, the ruined path where everybody jump when coing back from Meridia through the Djinnporter... right before the small ramp up are also good examples...
I can't say much for CiTW I hate that raid, the headless chicken running involved, all it's additionned retarded mechanism [ blown to death, death jail for a minute, mana sponging, spawning traps and respawning mob in cleared area, idiotic hand ]... not to mention the fact that I'm lagging behind because I'm afraid of falling.