open All Channels
seplocked EVE Information Portal
blankseplocked New Dev Blog: ing Lag: Module Lag - Why Not All Bugfixes Are A Good
 
This thread is older than 90 days and has been locked due to inactivity.


 
Pages: [1] 2 3 4 5 6 7

Author Topic

CCP Fallout

Posted - 2010.08.23 14:45:00 - [1]
 

We've all experienced this at one time or another: hit an F key, say an F word while waiting. What's that all about? CCP Veritas gives us the skinny in his new dev blog.

ivo
Drunken Pilots
Posted - 2010.08.23 14:54:00 - [2]
 

Its nice to know that this all powerful bug is finally getting squished.

Also, First, wow my first first Very Happy

Yuki Kulotsuki
Posted - 2010.08.23 14:57:00 - [3]
 

Nice face. Nothing like encountering "rare" exceptions regularly. This certainly sounds like a problem though. Finding the right balance of processing. I do not envy ya that job.

Zendoren
Aktaeon Industries
The Black Armada
Posted - 2010.08.23 15:03:00 - [4]
 

Nice follow up on the mod lag guys.....

If you keep this level of blog reporting.... I think you will gain many friends in the eve community!

Force Kon
Posted - 2010.08.23 15:10:00 - [5]
 

I hope this will finally show people that CCP are making steps to fix lag and not focusing all their efforts elsewhere.

TeaDaze
Agony Unleashed
Agony Empire
Posted - 2010.08.23 15:15:00 - [6]
 

Thanks for the work in tracking this down. We look forward to seeing the fix in action Smile

Qujulome
Amarr
Posted - 2010.08.23 15:15:00 - [7]
 

+1 for pictures Very Happy

Evelgrivion
Gunpoint Diplomacy
Posted - 2010.08.23 15:16:00 - [8]
 

Edited by: Evelgrivion on 23/08/2010 15:16:47
Unless I misread the blog, the means for actually fixing the issues aren't mentioned. Ive heard before that Dogma was written in Python, and that its the single largest bottleneck in EVE. What do you guys have in mind to bring Dogma up to speed? Will the relevant bits be refactored in C++ or are we going to be waiting for the Stackless Python Just In Time Compiler project to finish up, perhaps with some "minor" refactoring of the existing Stackless code?

Zirse
Minmatar
Brutor Tribe
Posted - 2010.08.23 15:22:00 - [9]
 

Thank you sir, for the read. Very Happy

Jorend Dohag
Posted - 2010.08.23 15:22:00 - [10]
 

You should look into GPGPU! It pays off.

Levitikon
Constructive Influence
Northern Associates.
Posted - 2010.08.23 15:25:00 - [11]
 

Oh, wow. So basically, great pre-dominion server performance had NOTHING TO DO WITH CCP, it was byproduct of bug that made your guns stuck, thereby lessening load on node system.


You really cannot, under no circumstances, yield Dogma subsystem - that will end up with warp scramblers not activating, guns not firing when they should and in effect totally ****ed up battles, where people can warp out/survive when they shouldn't, or where they die when they should lie (case: warpout command was issued 10 seconds after you got out of warp range, but since Domga was yelded, warp scrambling effect was never activated).


Yes, while you are reading it dear devs - there exists another CONFIRMED layer of this bug: With good enough lag, you can keep your target warpscrambled regardless of state of your module or DISTANCE.

You could* fit 7 warp scramblers on your ship, MWD to enemies (cool thing - module won't cycle, you will keep your MWD speed infinitively without using cap), activate scramblers on 7 enemy capital ships, fly 100KM away and enjoy perma-scrambled 7 with no risk for you.

I wonder if you could pull something like that with focused HIC and titan... I hope, dear developers, that you can see why no one wants to commit Titans into big laggy fights anymore.

Lin Xiao
Ars ex Discordia
Test Alliance Please Ignore
Posted - 2010.08.23 15:28:00 - [12]
 

Thanks for the update CCP Veritas.

This is an example of how to communicate and interact with your players. Keep it up.

Shamblingform
Gallente
Posted - 2010.08.23 15:29:00 - [13]
 

Easily first page! SurprisedSurprisedSurprisedShockedShockedShockedShocked

German Giggles! YARRRR!!YARRRR!!ughugh
R

CCP Veritas

Posted - 2010.08.23 15:29:00 - [14]
 

Originally by: Evelgrivion
What do you guys have in mind to bring Dogma up to speed? Will the relevant bits be refactored in C++ or are we going to be waiting for the Stackless Python Just In Time Compiler project to finish up, perhaps with some "minor" refactoring of the existing Stackless code?


I've been busy profiling it under some basic scenarios and there's some low hanging fruit to be plucked in there. There's significant algorithmic optimizations to be made as well, and those follow you around regardless of what language you're in. Language/platform shift isn't on the table until the algorithms are sound.

In other words, I believe Dogma is doing stupid things, and I intend to beat the stupid out of it before considering giving it rocket boots.

Jorend Dohag
Posted - 2010.08.23 15:29:00 - [15]
 

http://developer.download.nvidia.com/compute/cuda/3_1/toolkit/docs/NVIDIA_CUDA_C_ProgrammingGuide_3.1.pdf

Mashie Saldana
Minmatar
Veto Corp
Posted - 2010.08.23 15:30:00 - [16]
 

Originally by: Jorend Dohag
http://developer.download.nvidia.com/compute/cuda/3_1/toolkit/docs/NVIDIA_CUDA_C_ProgrammingGuide_3.1.pdf

Wrong tool for the job.

Levitikon
Constructive Influence
Northern Associates.
Posted - 2010.08.23 15:34:00 - [17]
 

Another useful tidbit of information - please include killing ships into your thin client simulations.

Shooting single titan is all fun and giddy, but in real battles people's ships go boom!. Associated with it - a lot of changing targets, loosing targeting locks, getting out of range, WARPING OUT AND IN (kind of important, as it's only real 'tank' players have against being primaried by enemy fleet and only tactical tool FCs have for implementing battle strategy).

Vuk Lau
4S Corporation
Morsus Mihi
Posted - 2010.08.23 15:34:00 - [18]
 

Edited by: Vuk Lau on 23/08/2010 15:34:29
BTW Video mentioned in this devblog CSM presented to CCP can be found on following link
Youtube link
Btw this is something called "manual gun cycling" and is used in fleet fights when node starts to die.

Thanx for keeping us in the loop about your progress.

Kikki Di'je
Lay Low
Posted - 2010.08.23 15:35:00 - [19]
 

I love you.

No, seriously, I love you.

I love you man. No, really.....I love you.

Hemmo Paskiainen
Gallente
Posted - 2010.08.23 15:36:00 - [20]
 

Edited by: Hemmo Paskiainen on 23/08/2010 15:36:38
Nice blog, module lag is one of the things that annoyed me im fleet fights

I love u tooYARRRR!!

CCP Veritas

Posted - 2010.08.23 15:39:00 - [21]
 

Originally by: Levitikon
Another useful tidbit of information - please include killing ships into your thin client simulations.


When doing this sort of analysis, it's best to start out by isolating exactly what you want to look at to remove any external influence. In the cases pictured, I wanted to observe module and drone behaviors, so that's exactly what I had going on - and little else.

I have done tests involving blowing up many ships (and little else). They were as fun as they were informative~

CCP Fallout

Posted - 2010.08.23 15:50:00 - [22]
 

Originally by: Vuk Lau
Edited by: Vuk Lau on 23/08/2010 15:34:29
BTW Video mentioned in this devblog CSM presented to CCP can be found on following link
Youtube link
Btw this is something called "manual gun cycling" and is used in fleet fights when node starts to die.

Thanx for keeping us in the loop about your progress.


Thanks. I ninja edited the link in :D

Vuk Lau
4S Corporation
Morsus Mihi
Posted - 2010.08.23 15:56:00 - [23]
 

Originally by: CCP Fallout
Originally by: Vuk Lau
Edited by: Vuk Lau on 23/08/2010 15:34:29
BTW Video mentioned in this devblog CSM presented to CCP can be found on following link
Youtube link
Btw this is something called "manual gun cycling" and is used in fleet fights when node starts to die.

Thanx for keeping us in the loop about your progress.


Thanks. I ninja edited the link in :D


Thank YOU :D

Indeterminacy
THORN Syndicate
BricK sQuAD.
Posted - 2010.08.23 15:57:00 - [24]
 

"we're tweaking the parameters of our existing infrastructure to improve performance."

That's basically the story amirite? Some specifics about the longer term solution would be useful.

Of course we all know why the devblog spam now. In the end they're good and of course it's good problems are being addressed. On the other hand, this seems elementary enough that it could have been tackled at any point in the last 6 months for little if any cost on CCPs part.

That's a bit depressing.

Lumy
Minmatar
Sebiestor Tribe
Posted - 2010.08.23 15:58:00 - [25]
 

Originally by: Indeterminacy
"we're tweaking the parameters of our existing infrastructure to improve performance."

That's basically the story amirite? Some specifics about the longer term solution would be useful.

Of course we all know why the devblog spam now. In the end they're good and of course it's good problems are being addressed. On the other hand, this seems elementary enough that it could have been tackled at any point in the last 6 months for little if any cost on CCPs part.

That's a bit depressing.

This reading thing, you aren't really good at it, right mate?

Ix Forres
Caldari
Righteous Chaps
Posted - 2010.08.23 16:04:00 - [26]
 

Great facial expression, good dev blog. How many major fixes have come about since the introduction of the thin client platform, and when was this made internally available? What were your methods of debugging EVE before it was available?


Originally by: Mashie Saldana
Originally by: Jorend Dohag
http://developer.download.nvidia.com/compute/cuda/3_1/toolkit/docs/NVIDIA_CUDA_C_ProgrammingGuide_3.1.pdf

Wrong tool for the job.


This. Whenever people go "ZOMG CUDA! OpenCL! GPU processing!" I just cry.

Yes, GPUs are fast at some tasks. However, for the majority of tasks (including the majority of EVE Online's server tasks) they are not suitable. They are applicable to very, very narrow fields- graphics rendering, massive processing of embarrassingly parallel (it's a technical term, fyi) tasks, and so on. None of which covers EVE's server side. And let's not forget the fact that TQ runs on blade servers; adding GPUs would require a shift to traditional server form factors which would require a massive investment to migrate, along with an increase in running costs, probably increases in other costs like maintenance, a requirement to reinvest in huge amounts of networking gear... and you're still using the wrong tool for the job.

Dierdra Vaal
Caldari
Veto.
Veto Corp
Posted - 2010.08.23 16:08:00 - [27]
 

Quote:
Essentially what we have to do is re-introduce yielding in the effect processing queue, only under our control instead of at the whim of some defect.


Does this mean the module lag will get better, or will it stay the same?

Aranial
Gallente
Empyrean Warriors
The Obsidian Front
Posted - 2010.08.23 16:09:00 - [28]
 

ah ha! MOAR Brain NOM NOM

Rakshasa Taisab
Caldari
Sane Industries Inc.
Posted - 2010.08.23 16:10:00 - [29]
 

Originally by: Ix Forres
This. Whenever people go "ZOMG CUDA! OpenCL! GPU processing!" I just cry.

And the specialized code that needs to go in to make it all work makes developers cry.

Dante Edmundo
Posted - 2010.08.23 16:18:00 - [30]
 

Edited by: Dante Edmundo on 23/08/2010 16:21:23

Excellent and insightful report.

Having written code myself for many years, I was often surprised how little emphasis would
be placed by project management on "thin testing clients" or even "automated testing". They
simply didn't want to spend the time up front to develop these valuable bug hunting tools.

Automated testing is especially powerful as it saves a good amount of developer time in the long run,
and re-checks older features automatically on each build - to make sure the code doesn't break something old
while fixing something new. It also is a good method of making sure your test itself is solid -
since an ad-hoc testing process can be buggy as well.

Good luck with your efforts. I think it benefits the community and CSM being able to read BLOG
posts of this detail. And shows you guys are sincerely attempting to work on issues that dog players
today - not just expansions for the future of Eve.

Dante




Pages: [1] 2 3 4 5 6 7

This thread is older than 90 days and has been locked due to inactivity.


 


The new forums are live

Please adjust your bookmarks to https://forums.eveonline.com

These forums are archived and read-only