| Pages: [1] 2 3 :: one page |
|
|
| Author |
Thread Statistics | Show CCP posts - 0 post(s) |
Sionn Klorgh
Minmatar
The Collective
Against ALL Authorities
   |
Posted - 2008.04.15 02:18:00 -
[1]
I've know 'lag' since the beginning. I believe CCP totally understands the issue also but there is some limitation it can not repair in the immediate future.
* Is it a hardware issue?
Lag tends to be ok when the fleet are in position but as more and more commands are pressed, it seems like a backlog of commands piles up giving this delay.
* Does the server sort command based on user latency?
That would sort out more commands faster by bias certain regional players.
My questions,
WHAT EXACTLY IS CAUSING LAG?
DOES CCP UNDERSTAND THE CAUSE EXACTLY?
IS THERE SOME FUNDAMENTAL LIMITATION IN TECHNOLOGY THAT WE ARE REACHING IN THIS GAME PREVENTING A FIX?
I appreciate everyone's opinion on the matter. |
DroneCommander
Gallente
|
Posted - 2008.04.15 02:20:00 -
[2]
People. |
Havohej
The Defias Brotherhood
ANTHRAX DEATH
|
Posted - 2008.04.15 02:21:00 -
[3]
/me would like to know more about what actually causes this as well, other than the whole 'Jita doesn't lag 'cause eve is broken, Jita lags because people keep going UNTIL it lags' routine. (same goes for fleets, btw) |
Liang Nuren
Black Sea Industries
Insurgency
|
Posted - 2008.04.15 02:23:00 -
[4]
Sigh, yes, they know. You really don't know how complicated the answer to your question is, or you wouldn't have asked it.
-Liang |
Sionn Klorgh
Minmatar
The Collective
Against ALL Authorities
|
Posted - 2008.04.15 02:27:00 -
[5]
I wonder if Jita has more hardware dedicated to it. The entire 'reinforcing' nodes...I wonder how well that works?
One should remember that the totality of fleet battle commands (and the fact that all the players are on the same grid) sounds like its a lot more demanding computationally. Does that mean its a mere CPU issue? Or is it a network speed issue? Or what?
I know we have some pros on this stuff on these boards whose opinions would be insightful. |
Crackzilla
The Shadow Order
SMASH Alliance
|
Posted - 2008.04.15 02:29:00 -
[6]
To quote a recently podded cliche:
Lag is people!
|
Lithalnas
Amarr
Headcrabs
|
Posted - 2008.04.15 02:31:00 -
[7]
lag is people, now in systems like motsu lag is isk farmers, ban the isk farmers to make regular mission runners happy. |
Sionn Klorgh
Minmatar
The Collective
Against ALL Authorities
|
Posted - 2008.04.15 02:32:00 -
[8]
I noticed the correlation between lag and numbers a very long time ago.
I'm after something more detailed (I'm a physicist in the real world). |
Havok Dryke
Golden Gavel Enterprises
The Cooperative
|
Posted - 2008.04.15 02:41:00 -
[9]
Edited by: Havok Dryke on 15/04/2008 02:41:20
The primary cause of lag are threads about the causes of lag.
--------------------------------------
EVE is a cold, harsh world, filled with people that would kill you without a second thought. The forums are even worse. |
Sionn Klorgh
Minmatar
The Collective
Against ALL Authorities
|
Posted - 2008.04.15 02:52:00 -
[10]
Glad to see everyone is as clueless about the issue as I.
|
|
|
Animin Mannja
|
Posted - 2008.04.15 02:56:00 -
[11]
Probably scalability issues.
All the neat things that make things interesting, for nonsensical exmple: shiny ships with role bonuses affecting fleet members overheating their turret tracking speed with specialised ammo against rigged ships piloted by pilots with boosters and implants and skills and all sorts of bling, all add to the complexity of situations and thus the computational power required to resolve them.
Of course, Jita's probably also collapsing under the weight of Market orders being updated several hundred times a second, and making sure that you can actually buy what you can see and that it hasn't changed in the last 30th of a second.
All this sort of stuff taxes resources to the limit. Seems we've been hitting the limits more often recently though, new people trying out new features, etc.
Basically, the system isn't scaling as well as CCP would hope.
The short term fix is "Throw more power at it!" which is what is frequently spouted around here. That just extends the limit a bit, and is really impractical to be done on a large scale - dedicated nodes for each system just means 2/3rds of the shiny expensive equipment sitting idle as they never get close to the limit.
The long term fix is to improve scalability by re-working some of the base code and infrastructure, compartmentalising it so the workload can be distributed better across multiple servers. This of course takes a lot of time and effort, and may not be possible in some cases - some things can't be separated.
|
Hannobaal
Gallente
Shadow Forces Inc.
|
Posted - 2008.04.15 02:57:00 -
[12]
Edited by: Hannobaal on 15/04/2008 02:57:07
 Originally by: Sionn Klorgh
WHAT EXACTLY IS CAUSING LAG?
Too many people in the same place doing too many things. What else could it possibly be?
Quote:
DOES CCP UNDERSTAND THE CAUSE EXACTLY?
I think it's a pretty safe assumption to say that they understand the causes better and in more detail than any of us possibly could.
Quote:
IS THERE SOME FUNDAMENTAL LIMITATION IN TECHNOLOGY THAT WE ARE REACHING IN THIS GAME PREVENTING A FIX?
Yeah.
|
Jovian Strain
The Collective
Against ALL Authorities
|
Posted - 2008.04.15 02:59:00 -
[13]
Quote:
Too many people in the same place doing too many things. What else could it possibly be?
No doubt. Is that a CPU issue or a network issue or what?
|
Animin Mannja
|
Posted - 2008.04.15 03:18:00 -
[14]
Originally by: Jovian Strain
Quote:
Too many people in the same place doing too many things. What else could it possibly be?
No doubt. Is that a CPU issue or a network issue or what?
Probably both. Maybe not at the same time. I'm sure there are instances where the server can't keep up with computational tasks, and others where it can't spit out info to clients fast enough.
I'd expect there are many different events affecting many different aspects of the servers, all of which would be seen as just 'lag'.
|
Sionn Klorgh
Minmatar
The Collective
Against ALL Authorities
|
Posted - 2008.04.15 03:21:00 -
[15]
Quote:
Probably scalability issues.
All the neat things that make things interesting, for nonsensical exmple: shiny ships with role bonuses affecting fleet members overheating their turret tracking speed with specialised ammo against rigged ships piloted by pilots with boosters and implants and skills and all sorts of bling, all add to the complexity of situations and thus the computational power required to resolve them.
Of course, Jita's probably also collapsing under the weight of Market orders being updated several hundred times a second, and making sure that you can actually buy what you can see and that it hasn't changed in the last 30th of a second.
All this sort of stuff taxes resources to the limit. Seems we've been hitting the limits more often recently though, new people trying out new features, etc.
Basically, the system isn't scaling as well as CCP would hope.
The short term fix is "Throw more power at it!" which is what is frequently spouted around here. That just extends the limit a bit, and is really impractical to be done on a large scale - dedicated nodes for each system just means 2/3rds of the shiny expensive equipment sitting idle as they never get close to the limit.
The long term fix is to improve scalability by re-working some of the base code and infrastructure, compartmentalising it so the workload can be distributed better across multiple servers. This of course takes a lot of time and effort, and may not be possible in some cases - some things can't be separated.
So ultimately, your view is that its a CPU issue. A typical CPU gets so many node to handle. Through 500 people fleet battle on that node and it just runs out of steam.
- I have heard that putting more cpu onto a node helps (reinforcing a node).
- Pity this can't be dynamic based on the number of people in a system. I believe 'reinforcing' can only be done at downtime.
|
Animin Mannja
|
Posted - 2008.04.15 03:33:00 -
[16]
Originally by: Sionn Klorgh
So ultimately, your view is that its a CPU issue. A typical CPU gets so many node to handle. Through 500 people fleet battle on that node and it just runs out of steam.
- I have heard that putting more cpu onto a node helps (reinforcing a node).
- Pity this can't be dynamic based on the number of people in a system. I believe 'reinforcing' can only be done at downtime.
IMO, in a lot of instances it's probably a CPU type issue. Might not be CPU per se, might be RAM, bus/disk I/O, etc, etc. Basically it can't keep up.
I suspect 'reinforcing a node' is basically putting the particular systems on their own heavy-duty server, which would only be able to be done at downtime without interrupting anything, so is unable to be done dynamically with the current infrastructure.
However I believe Jita is already on a permanent reinforced node, so it's obviously not a cure-all. I don't think there can be a cure-all with the current infrastructure, which just means lots of threads like this until they finally finish Infiniband or whatever they've got in the pipeline at the moment.
|
Oldin Kinrod
|
Posted - 2008.04.15 03:34:00 -
[17]
A couple of dev blogs about lag:
The EVE Cluster
Optimizations, Optimizations, Optimizations (Has animation explaining it)
People stealing your wrecks? Players intruding your missions? |
Akita T
Caldari Navy Volunteer Task Force
|
Posted - 2008.04.15 04:04:00 -
[18]
Client-side lag = not enough computing power to render the game (doesn't affect much except your FPS)
Network lag = not enough bandwidth to transfer all data, or a bad response time (hardly the case these days, but can happen)
SERVER-SIDE lag = not enough CPU to compute all user actions (the main cause of "lag" in this game)
The answer is to simply just throw more computing power at it.
Ok, it's a SIMPLE answer, but it's not an EASY answer.
The big problem with this answer is economic concerns... yeah, you COULD have each solar system on a machine that could handle a couple of thousands of people doing their thing, however that machine might cost more than the whole yearly INCOME of CCP.
So, you have to prioritize.
Meanwhile, a project is underway to try and change the way server-side simulations are handled, from "several solar systems on a certain node" (like it is now, with a few nodes running single solar systems, like Jita, for instance) to a more fluid approach, where SEVERAL nodes could be used to handle a single solar system, and "on the fly" rerouting of server hardware where needed.
Sadly, this is not a project you do overnight, and it might be YEARS before that finally happens, IF it happens at all.
1|2|3|4|5. |
Rawr Cristina
Caldari
Naqam
|
Posted - 2008.04.15 04:12:00 -
[19]
Originally by: Hannobaal
Too many people in the same place doing too many things.
with no mechanics, other than that of lag, to persuade people to avoid such places.
...
|
Pithecanthropus
|
Posted - 2008.04.15 04:13:00 -
[20]
Bad programming
Lack of server dedication
Never made preparations for game growth
Adding content before adding quality
Too much money makes them lazier
All of the above leads to lag. Players don't cause lag, we're victims of it due to the problems CCP refuses to fix.
|
|
|
Crackzilla
The Shadow Order
SMASH Alliance
|
Posted - 2008.04.15 04:21:00 -
[21]
Originally by: Pithecanthropus
Bad programming
Lack of server dedication
Never made preparations for game growth
Adding content before adding quality
Too much money makes them lazier
Generally its a business issue. No sane business person would sign off on developing a massively scalable system to be deployed years from the planning stage without customers. Now they've customers and a business need.
Some groups are tasked on performance but not all can be. Its just a matter of time. On the other hand the roi case may say to keep things as is and do nothing. Who knows.
|
Akita T
Caldari Navy Volunteer Task Force
|
Posted - 2008.04.15 04:21:00 -
[22]
Originally by: Pithecanthropus
Bad programming
Lack of server dedication
Never made preparations for game growth
Adding content before adding quality
Too much money makes them lazier
All of the above leads to lag. Players don't cause lag, we're victims of it due to the problems CCP refuses to fix.
Would that be the reason that "once upon a time" 50vs50 were already not just lagfests, but freezefests, yet now you can even have 200vs200 and be "only" lagfests ?
Or is that the reason why Yulai, then Jita used to croak at 200-300-ish in local, but now you can take a stroll at 600 in local and still do your thing ? Or that at 800, it's still barely playable ? And only near 1k it finally croaks ?
Yup, there was absolutely no progress, no sir...
1|2|3|4|5. |
Pithecanthropus
|
Posted - 2008.04.15 04:27:00 -
[23]
Edited by: Pithecanthropus on 15/04/2008 04:30:16
Originally by: Akita T
Yup, there was absolutely no progress, no sir...
If progress led to where they are at... progress can lead to better. Thus, they are still bad, lacking, not making adjustments, and turning their heads from the problems.
I refuse to be at fault for lag, when the game THEY created molds us to do what we do. If you make a narrow tunnel and tell 30,000 people to run thru it, you'll get stuck... same goes for this game. CCP made it, CCP is at fault for opening the doors and letting us get stuck. CCP is at fault for putting out more quantity and less quality. I play a game, I play with the mechanics they gave me... why should lag be MY fault?
CCP has the accountability... you fail.
EDIT: and to be honest... I'm the one that has to adjust the game to make it play less lag free. Cliking off effects, no sound, no high graphics content... so yea... I'm doing what I can do. Taint my fault if I fight 200 vs 200... CCP wants us to. They just aren't prepared for it.
|
Sionn Klorgh
Minmatar
The Collective
Against ALL Authorities
|
Posted - 2008.04.15 05:05:00 -
[24]
Quote:
Meanwhile, a project is underway to try and change the way server-side simulations are handled, from "several solar systems on a certain node" (like it is now, with a few nodes running single solar systems, like Jita, for instance) to a more fluid approach, where SEVERAL nodes could be used to handle a single solar system, and "on the fly" rerouting of server hardware where needed.
So a CPU issue. The nodes are spread out over CPUs and these CPUs are not currently coded to dynamically redistribute load, so we get lag. Extra hardware thus will not do as much as some fundamental re-writing of the source code.
|
LaVista Vista
Conservative Shenanigans Party
|
Posted - 2008.04.15 05:12:00 -
[25]
We REALLY need a sticky about the lag. I have explained this in 6 threads the last week.
The issue is that a system contains certain services. The market, space, etc. etc., and these are forced to be on 1 core(not a cpu, just a core), because the cluster talks to nodes over ethernet connection.
Therefor, the cluster can't scale by redistributing the load between more nodes, due to the ethernet connection having a large latency.
CCP is implementing infiniband to the cluster this year, in order to make it possible, and thus killing the lag-monster.
|
Zeba
Minmatar
Pator Tech School
|
Posted - 2008.04.15 05:46:00 -
[26]
There are three types of lag. One is market hub lag with hundreds of trade alts all endlessly refreshing the market interface to gain the .01 isk advantage. Two is empire mission hub lag with thousands of players spamming the same Caldari Command agents thinking they are the only ones to give out faction missiles and CNRs. Three is 0.0 lag due to having whole regions running on one cpu so when a fleet fight happens and some random nub accesses the market the node crashes.
One and Two are totally player driven and CCP will never be able to do anything about it and Three is simply because the people who do fleet fights only make up about 15% of the playerbase and so get the comparative equivilent of server resorces.
Thats about it. 
Why not. |
Robdon
|
Posted - 2008.04.15 05:53:00 -
[27]
Bad Coding.
Jugging by the amount of bugs and CCPs inability to fix them in any kind of quick manner, I would say that the state of the code must be appalling.
I've worked with computing for many many years, and most coders donĘt have a clue about performance, or even consider it.
You have to have a couple of guys who dedicate themselves to this area and review / analyse ALL code that gets into the system.
ItĘs a completely different thought process, that CCP just doesnĘt seem to have.
Now obviously I havenĘt seen the code, but things like this....
Fleet battles, the first thing you do is lock a target and fire all your weapons.
This involves clicking on maybe 6-7 (13 or more for ECM boats) modules to activate all your weapons, and thatĘs 7 transactions to the server that need to be processed.
There should be an 'activate all' button, that as a user you can setup and decide what modules get activated with this button, then if this is pressed 1 transaction is sent to the server, instead of 7.
This is just one example that I can see 'should' help.
It might sound harsh, but since trinity and the 'Need for speed', this game has just got worse. So what ever it is their coders are doing itĘs not what is needed. They need to get fresh dedicated coders who know what they are doing in this area, because CLEARY the current ones donĘt know what to do.
|
Zeba
Minmatar
Pator Tech School
|
Posted - 2008.04.15 05:55:00 -
[28]
Originally by: Robdon
lulz ima idjut
Bad reasoning.
Why not. |
CHAOS100
Black Plague.
RAZOR Alliance
|
Posted - 2008.04.15 06:04:00 -
[29]
Despite what most people think i'm pretty sure it is not a hardware issue. They have a very powerful computer cluster, which they could play pong quite nicely on. You could have the whole earth as a computer and it will still only be as fast and as good as your programming lets you be. Let's be honest, they developed the primary engine of the game back in 2002 or so? Sure they made a new graphics engine, but that is good for the client not the server. New network coding would probably be very difficult seeing as how it is kind of like trying to remake the foundation of a house. But CCP is just putting more layers on a broken foundation.
I remember distinctly that around the end of Cold War client and the entrance into RMR that lag started to get bad. I am not talking about player increase. I remember that we could have good sized battles right before the patch to RMR, with minimal lag. Then RMR hit and a battle with say 200 people was nearly unplayable.
Check out kirkaphoenix's video 'Where have all the good fix gone' which was frapsed right after RMR. The lag in that video in unbearable. The player count did not rise that significantly after the patch to make a major difference. This leads me to believe that whatever code and 'content' that patch introduced suddenly made the servers run much less efficient, which they have been continually trying to fix by improving the server cluster.
So it is probably a coding/software issue, which CCP is probably reluctant to fix due to the large amount of time and effort required to re-code it, and would add no new visible value to the game that say... walking in station would.
--------------
|
Inertial
The Python Cartel
|
Posted - 2008.04.15 06:41:00 -
[30]
Lag is caused by the server/network/client needing to compute more info than it can. Just like a Ibis can't break a passive drakes tank, a Intel 4004 can't handle all the instructions EVE gives it in time, before it is given a new set of instructions.
The problem with lag in eve is: Whenever there are a person in the system, the server needs to know;
Where the person is
What he is in
The condition of his ship
The speed of the ship
Modules activated
etc. etc. etc.
All of this adds up, and when you have 600 people in a system it needs to know this for every one of them. This is especially bad in case of a fleet battle, where information about damage taken and ships destroyed is constantly sent, or Jita, where market information is constantly updated.
|
|
|
|
|
| |
|
| Pages: [1] 2 3 :: one page |
| First page | Previous page | Next page | Last page |