Pages: [1] 2 :: one page |
|
Author |
Thread Statistics | Show CCP posts - 1 post(s) |
Recluse Viramor
Chosen Path Center for Disease Creation
|
Posted - 2006.08.07 00:22:00 -
[1]
Edited by: Recluse Viramor on 07/08/2006 00:23:09 The one issue that cripples EVE more than any other is lag. CCP continues to urge and create more gameplay elements that create the need for ever and ever larger fleet engagements yet the server architecture currently fails miserably at supporting it.
Currently the servers dynamically allocate server resources on a week to week basis based on system numbers, this needs to change.
A system needs to be developed that can allocate server resources on the fly when they are needed most. I am by no means someone with any knowledge of servers or how this would work, but lag is the single largest issue in EVE right now. The inability to play this game as intended without losing ships and billions of assets due to lag beyond your control is troubling.
Something must be done. Whether it be a reworking of the server architecture or a change in direction of EVE.
|
Taaketa Frist
Information Science Security Knights Of the Southerncross
|
Posted - 2006.08.07 00:30:00 -
[2]
with the current server side software/hardware setup it simply is not possible to reallocate server resources on the fly and each time it must be done during downtime.
Remember that when servers are reallocated to help focus on one system that means another node will then have to take on the load of the rest of that node's abilities.
Also note that if enough systems where to become busy enough and more nodes got allocated to support those systems, even sparsely populated areas of Eve could suffer from lag as more nodes become focused on one systems and fewer nodes then become extreme generallists.
For this reason the server only reallocate nodes for a specific system if its had for example more than 300 players in for 5 days running.
Ovuer mentioned something about the Dragon server software being applied to TQ but I don't know what the time scale is for that. Apprantly its a more upgraded form of the allocation system that is present on TQ at the moment. --------------
Dang nabit |
Recluse Viramor
Chosen Path Center for Disease Creation
|
Posted - 2006.08.07 00:36:00 -
[3]
Edited by: Recluse Viramor on 07/08/2006 00:40:02 Edited by: Recluse Viramor on 07/08/2006 00:37:13 Would it not be feasible to have reserver server resources specifically for these situations as they arise?
Epic fleet engagements are becoming ever more common. In the past it was only battleships that were lost due to lag. Now its dreads/carriers and perhaps even larger.
Honestly, motherships and titans are in game and can be used, but who would ever think of using them with the current state of lag? The server simply does not support the type of fleet engagement that would call for the use of a titan.
The current architecture is fundamentally flawed.
Flawed Game mechanics are currently dictating the poltical boundaries, and no better an example than this announcement by LV.
|
Taaketa Frist
Information Science Security Knights Of the Southerncross
|
Posted - 2006.08.07 00:52:00 -
[4]
Edited by: Taaketa Frist on 07/08/2006 00:52:24 IF you mean by informing Gms about large fleet activity movements or some such before you, what garuntee that all 300 players will ACTUALLY engage in the system you requested the node to be reinforced?
Combat can happen anywhere and combat may spill over into other systems. People are hard to rely on at the best of times... multiply that by 300 people its just no tenable.
edit: ovuer made a post not so long about this... gonna poke around. --------------
Dang nabit |
Recluse Viramor
Chosen Path Center for Disease Creation
|
Posted - 2006.08.07 01:02:00 -
[5]
Originally by: Taaketa Frist Edited by: Taaketa Frist on 07/08/2006 00:52:24 IF you mean by informing Gms about large fleet activity movements or some such before you, what garuntee that all 300 players will ACTUALLY engage in the system you requested the node to be reinforced?
Combat can happen anywhere and combat may spill over into other systems. People are hard to rely on at the best of times... multiply that by 300 people its just no tenable.
edit: ovuer made a post not so long about this... gonna poke around.
Not not by people at all, but by the servers themselves self-allocating based on system numbers. An architecture that can allocate resources fluidly is what I'm getting at.
|
Lintaka
|
Posted - 2006.08.07 01:19:00 -
[6]
Originally by: Taaketa Frist
IF you mean by informing Gms about large fleet activity movements or some such before you, what garuntee that all 300 players will ACTUALLY engage in the system you requested the node to be reinforced?
Still doesnĘt work, there has been a fleet battle recently where the GM's were apparently informed a week before that there would be a large scale battle happening, they either didnĘt listen, as the battle was still incredibly laggy and the op was scrapped in the end due to high losses, even though if things ran smoothly it should have been a success without a problem.
Or flip the coin they did listen, and any changes they made blatantly did not work..
|
Deva Blackfire
DAB RAZOR Alliance
|
Posted - 2006.08.07 02:01:00 -
[7]
Lol... just check XZH. It was going for over a month with over 200 players in combat every day (usually closer to 300 with 400-500 in peak times). One month and we didnt get any support - in the end node crashed and for next 2+ days it was impossible to play without lag in that area (even after numbers dropped to 50 or so).
|
Arlenik Emmanouelik
|
Posted - 2006.08.07 03:54:00 -
[8]
AFAIK, yet I could be wrong on this since this might have changed by now, more than one solar system can be assigned to a node. However, no more than one node can be assigned to a solar system. This is the main limiting factor of TQ design.
|
Dao 2
The Scope
|
Posted - 2006.08.07 03:59:00 -
[9]
it is true though :| eves biggest problem is lag, bar none ;p ------------------------------------------------ NEWLY ADDED ON 1/19 (though applies to all posts before ;p)
the usual "I don't represent my corp or alliance" and stuffs like that
Also the gal |
Willis Drummond
Lordless Unbrella Alliance
|
Posted - 2006.08.07 04:52:00 -
[10]
Either way the clear answer here is that we need more python code.
Awesome Post# 677886 |
|
Emily Spankratchet
Minmatar Pragmatics
|
Posted - 2006.08.07 06:56:00 -
[11]
Originally by: Recluse Viramor An architecture that can allocate resources fluidly is what I'm getting at.
That's what we need, yes.
Unfortunately, I get the impression that it would involve a ground-up redesign of the entire cluster architecture. So suggesting it to the Dev team would probably make them go out back and shoot themselves.
|
Splagada
Minmatar
|
Posted - 2006.08.07 08:39:00 -
[12]
i never saw a MMO without lag
with the years it gets way better tho
remember UO release?
THAT was lag -
Looking for technetium delivery contracts
|
WeetBix
ANZAC ALLIANCE Miners With Attitude
|
Posted - 2006.08.07 09:05:00 -
[13]
Originally by: Splagada i never saw a MMO without lag
I think the problem is that in most other games lag is no big deal. In WoW you die, you respawn and go on your merry way. In BF2 you die, respawn and run back into the battle. At worst you go 'aww snap'. In Eve a bit of lag can have much more devastating consequences, so I understand why people don't want to lose months worth of work just because of a lag hit (and why it's such a big topic).
Originally by: Splagada
remember UO release?
THAT was lag
LMAO
----------- If I see one more 'FIRST!' post, I'm gonna snap. |
Batleth Bladed
|
Posted - 2006.08.07 09:14:00 -
[14]
signed :)
|
Jessica McDaniel
|
Posted - 2006.08.07 09:15:00 -
[15]
singed, god the lag is FKin awful i lost my raven to lag today ::::::: i'm ****ed:::::: |
Taaketa Frist
Information Science Security Knights Of the Southerncross
|
Posted - 2006.08.07 09:28:00 -
[16]
Edited by: Taaketa Frist on 07/08/2006 09:28:40 I can't find that post. But it was fully explained by a Dev was it was not possible.
The XZH was also explained that manually switching over nodes to focus on one solar system was potentially more destructive than letting the automatic process do it.
And as I said. What GARUNTEE can you give CCP that you will actually engage a fight in a specific solar system? Fleets can move you know?
edit: don't fly what you can't afford to lose is the motto for anyone who says they lost their life fortunes. --------------
Dang nabit |
Jenny Spitfire
Caldari LoneStar Industries Veritas Immortalis
|
Posted - 2006.08.07 09:39:00 -
[17]
Semi-grid instancing? ---------------- Cruelty is God's way of showing kindness.
|
Taaketa Frist
Information Science Security Knights Of the Southerncross
|
Posted - 2006.08.07 09:49:00 -
[18]
Originally by: Jenny Spitfire Semi-grid instancing?
You make the front page cry... CRY I tell you.
Quote: 25k Peak Concurrent User barrier broken Yesterday evening at 20:48 a new PCU milestone was reached when 25.665 EVE Players were logged in at the same time. This record follows a highly successful hardware upgrade of CCP's server cluster in London allowing CCP to remain true to its regiment of one unsharded virtual world with no instancing.
Why do you do it Jenny? Why must you make it cry. --------------
Dang nabit |
FireFoxx80
Caldari E X O D U S Imperial Republic Of the North
|
Posted - 2006.08.07 10:02:00 -
[19]
Originally by: Splagada i never saw a MMO without lag
with the years it gets way better tho
remember UO release?
THAT was lag
*wipes a tear from his eye*
lag-walking for the win!
|
Jenny Spitfire
Caldari LoneStar Industries Veritas Immortalis
|
Posted - 2006.08.07 10:04:00 -
[20]
Originally by: Taaketa Frist
Originally by: Jenny Spitfire Semi-grid instancing?
You make the front page cry... CRY I tell you.
Quote: 25k Peak Concurrent User barrier broken Yesterday evening at 20:48 a new PCU milestone was reached when 25.665 EVE Players were logged in at the same time. This record follows a highly successful hardware upgrade of CCP's server cluster in London allowing CCP to remain true to its regiment of one unsharded virtual world with no instancing.
Why do you do it Jenny? Why must you make it cry.
It isnt the kind of instancing you are thinking about. If I would to explain, it would be a geek talk. If you get it, you get it. If you dont, never mind. Another time when I am less busy. ---------------- Cruelty is God's way of showing kindness.
|
|
Zaldiri
Caldari Automated Industries
|
Posted - 2006.08.07 10:08:00 -
[21]
Its probably the best solution but would require an entire ground up redesign of the eve cluster. Would take years.
----------------------------------------------- Admiral of King Frieza's Super Saiyan fleet.
|
Recluse Viramor
Chosen Path Center for Disease Creation
|
Posted - 2006.08.07 10:43:00 -
[22]
This issue will just become larger and larger, perhaps if the devs cannot do anything an alternative would be to manually allocate an entire node to each and every system with an outpost or conquerable station... It's an extreme way of solving the issue but it could quite possibly solve the great majority of issues.
|
Taaketa Frist
Information Science Security Knights Of the Southerncross
|
Posted - 2006.08.07 10:54:00 -
[23]
Originally by: Recluse Viramor This issue will just become larger and larger, perhaps if the devs cannot do anything an alternative would be to manually allocate an entire node to each and every system with an outpost or conquerable station... It's an extreme way of solving the issue but it could quite possibly solve the great majority of issues.
Please read my posts. I already stated that manually assigning nodes according to the devs is more detrimental to the server stability than the automated process. --------------
Dang nabit |
|
Ivan Kirilenkov
Forum Moderator Interstellar Services Department
|
Posted - 2006.08.07 11:09:00 -
[24]
There's been a few post from Oveur on this matter; see this, this and this
eve-crc.net | forum rules |[email protected] |
|
Taaketa Frist
Information Science Security Knights Of the Southerncross
|
Posted - 2006.08.07 11:14:00 -
[25]
Originally by: Ivan Kirilenkov There's been a few post from Oveur on this matter; see this, this and this
And those were the posts I Was talking about. Couldn't blooming find them. --------------
Dang nabit |
Fubear
|
Posted - 2006.08.07 11:28:00 -
[26]
Edited by: Fubear on 07/08/2006 11:31:58 The load balancing needs to be re-written so that it is fully dynamic.
The biggest problem with the laod balancing is that it is currently impossible to have any kind of large engagements unless both sides sit in the same system for a week.
There is no way to know that a particular system will need increased server resources until it actualy needs them. The server should not have to 'wait until downtime' to reallocate resources, because by then it is far too late to help anyone.
The current balancing routine also doesn't work because it has takes into account the overall population of a system over 23 hours.* Even though there may be 800 people in local at one time, the balancing routine may not assign that system the resources to handle 800 people because the 50 servers assigned to that system may be sitting idle for the rest of the day once the battle is over and everyone leaves. To actually get the server performance you need for large scale battles, you need a large amount of people in the area all day every day for a few days straight.
The current load balancing system is broken, it is impossible to get the server resources for a large battle unless you stick around for a few days, and even then you don't get the resources to handle 'peak time' because of the wasted server resources during 'off time'.
A dynamic load balancing system is required, otherwise we will never be able to have fleet battles outside of 'hub' systems.
* This information was provided by Oveur when he dropped into XZH local last week.
EDIT: This info is also confirmed in the ISD link above. Maybe the new code branch will help things, but it is still not a proper solution.
|
Rikkard Strofeldt
Gallente SniggWaffe
|
Posted - 2006.08.07 11:53:00 -
[27]
VMware's ESX server might be handy for this. Each system gets its own virtual machine, and the VMotion software will start moving those machines on the fly while they are still running to let the busy machines get more of the processor. Still can't have a system on more than one node, but at least a system can have the node to itself without waiting for a downtime.
This is just me, but I wouldn't have Jita, Oursalert, or the other hubs inside VMs, they can have on the node directly.
If you're really lucky, this could make daily downtime become weekly downtime ;)
Also, not sure if it would be cheaper for CCP to deploy this, or create their own in-house version customised to EVE.
Besides, the devs have probably looked at this. I know I started drolling at the possibilities for using this where I work, and its going in Soon(tm) --
Descending into madness. |
Yarek Balear
Black Omega Security E.R.A
|
Posted - 2006.08.07 12:05:00 -
[28]
Well, I might be wrong, but the current lag issues don't appear to be completely node related to me. There might be an element of it in large engagements or busy systems but recently I've been experiencing a very high level of client side lag... I experience it no matter what system I'm in and it doesn't alleviate until I reboot the client (which appears to be using 97% CPU until I bounce it. Once I've bounced the client it appears to alleviate my problem for a while - hence the indication that what I experience is more client side than server side.
|
Eternal Fury
Caldari Brotherhood of Light
|
Posted - 2006.08.07 12:57:00 -
[29]
Unfortunately, I've never been in a battle with more then 25 player run ships. But as my corp continues to grow, the engagements will get larger and larger.
Real dynamic load balanceing may not be possible with the current code, but I hope the devs have thought to put this in the new client they are makeing for Vista/dx10. We may have to wait till then for this to be implemented.
Personally, the idea that nodes are assigned on a average of the past week seems useless.
some systems in 0.0 only have 5-10 people in them scattered throughout the day, untill some other alliance comes in with a fleet of 10-20 dreads, and a few hundred other combatants, and then you bring a few hundred of your alliance mates, and BOOM, over a 6hr period, you may go from 5 people in local to 400-500 in local. POS's go pop, ships get destroyed, and a day later there is no fighting in that system.
With a truly dynamic load balance, that system would get more and more resources as the day progressed. Slowly being given more and more horsepower, to keep that system lag free. And as soon as the battles are over, those resources would be free instead of being tied up till downtime, or untill the next day or 3 when things are "balanced".
We just broke the concurrent user record again yesterday. 26,827. This will keep increaseing as the subscriber's go up and up.
And come the vista/dx10 update, we all know the CCP marketing team will wind up and fire out a new round of adverts extoleing the virtues of the new code, and we'll get MORE new people.
I truely hope they get this resolved, as I'm here for the long haul. I've only reached 6.5 mill SP, but I plan on being here for a long time. So I'd really like to hear that CCP has a solution to this.
Brotherhood of Light. Small Corp, Big Fun. Wanna join? |
Virida
Mindstar Technology United Confederation of Corporations
|
Posted - 2006.08.07 14:29:00 -
[30]
Anarchy online got dynamic allocation of server resources, and, all who has played it know exacly how different that is from EVE(worse).
|
|
|
|
|
Pages: [1] 2 :: one page |
First page | Previous page | Next page | Last page |