Pages: 1 [2] 3 :: one page |
|
Author |
Thread Statistics | Show CCP posts - 0 post(s) |
bloody johnroberts
Caldari Shiva Morsus Mihi
|
Posted - 2010.01.25 22:49:00 -
[31]
you just have to watch the great fanfest vid on dominion and listen to there great statements on how it would change the way eve plays lol
yep we see it nice planets lol but game is unplayerble
everyone knows its the patch that has caused this but ccp wont agree because its a bad thing to say we screwed up as they are using dominion as a selling tool lol sandbox more like ashtray
|
Kinsy
Caldari Viper Squad Triumvirate.
|
Posted - 2010.01.25 23:18:00 -
[32]
roll it back to before dominion.
noone likes the new sov system anyway
PROBLEM ****ING SOLVED
Inappropriate content for signature, please change. ~Weatherman |
Dr Ngo
Amarr Shut Up And Play WE FORM VOLTRON
|
Posted - 2010.01.25 23:22:00 -
[33]
Originally by: Kinsy roll it back to before dominion.
noone likes the new sov system anyway
PROBLEM ****ING SOLVED
Keep all the suppercap changes though
|
Twigand Berries
Caldari GoonFleet GoonSwarm
|
Posted - 2010.01.25 23:35:00 -
[34]
Originally by: General Windypops Surely we could at least save a few terriflops of data by automacroing
10 print "fofofofo" 20 print "ur momma so" 30 goto 10
I know it's possible because Plague Black and Peter Quaersareana already operate on the same basic principle.
I don't know much about programming, etc, but it seems to sure make us laugh when people think local has anything to do with the lag.
|
Taedrin
Gallente The Green Cross DEFI4NT
|
Posted - 2010.01.25 23:51:00 -
[35]
Originally by: Garulf Hirgon Throw money at the problem: if they can't code for performance, buy better hardware. CCP should be swimming in cash...they're bringing in like $4 million USD a month and they can't have THAT much for overhead and operating expenses...
Because CPU clockspeeds are totally continuing to rise at an exponential rate.
OH wait... That's right. CPU clock speeds haven't increased since the pentium 4. Multiple cores? Don't make me laugh. Those can only increase performance in multi-threaded applications. EVE is single threaded. That means that each solar system in EVE can be handled by AT MOST 1 CPU core.
Sorry to burst your bubble there, but it is impossible to "buy better hardware", because better hardware just simply doesn't exist anymore. The fastest server CPU I can see on newegg is clocked at 3.3 GHz (costing 1,000 USD for the cheapest one). You will note that Tranquility's hardware is ALSO clocked at 3.3 GHz. CCP is already using the best hardware commercially available. ---------- There is always a choice. The choice might not be easy, nor simple, nor the options be what you desire - but, nevertheless, the choice is there to be made. |
Charlemeign
Gallente BESTIAL CARNAGE DEFI4NT
|
Posted - 2010.01.26 00:13:00 -
[36]
Edited by: Charlemeign on 26/01/2010 00:12:50
Originally by: Taedrin
Originally by: Garulf Hirgon Throw money at the problem: if they can't code for performance, buy better hardware. CCP should be swimming in cash...they're bringing in like $4 million USD a month and they can't have THAT much for overhead and operating expenses...
Because CPU clockspeeds are totally continuing to rise at an exponential rate.
OH wait... That's right. CPU clock speeds haven't increased since the pentium 4. Multiple cores? Don't make me laugh. Those can only increase performance in multi-threaded applications. EVE is single threaded. That means that each solar system in EVE can be handled by AT MOST 1 CPU core.
Sorry to burst your bubble there, but it is impossible to "buy better hardware", because better hardware just simply doesn't exist anymore. The fastest server CPU I can see on newegg is clocked at 3.3 GHz (costing 1,000 USD for the cheapest one). You will note that Tranquility's hardware is ALSO clocked at 3.3 GHz. CCP is already using the best hardware commercially available.
This is where the Black Market & NSA connections come in to play.
|
Xiaodown
Caldari Guiding Hand Social Club Dystopia Alliance
|
Posted - 2010.01.26 01:41:00 -
[37]
Originally by: Taedrin
Originally by: Garulf Hirgon Throw money at the problem: if they can't code for performance, buy better hardware. CCP should be swimming in cash...they're bringing in like $4 million USD a month and they can't have THAT much for overhead and operating expenses...
Because CPU clockspeeds are totally continuing to rise at an exponential rate.
OH wait... That's right. CPU clock speeds haven't increased since the pentium 4. Multiple cores? Don't make me laugh. Those can only increase performance in multi-threaded applications. EVE is single threaded. That means that each solar system in EVE can be handled by AT MOST 1 CPU core.
Sorry to burst your bubble there, but it is impossible to "buy better hardware", because better hardware just simply doesn't exist anymore. The fastest server CPU I can see on newegg is clocked at 3.3 GHz (costing 1,000 USD for the cheapest one). You will note that Tranquility's hardware is ALSO clocked at 3.3 GHz. CCP is already using the best hardware commercially available.
Dude, what?
Yes, all of eve is running on one single core 3.3Ghz proc.
There is NO WAY that serialized mathematical calculations could ever be multithreaded.
Are you seriously just spouting stuff you read on digg 5 years ago? --
|
Batolemaeus
Caldari Free-Space-Ranger Morsus Mihi
|
Posted - 2010.01.26 02:30:00 -
[38]
Originally by: Xiaodown
Yes, all of eve is running on one single core 3.3Ghz proc.
There is NO WAY that serialized mathematical calculations could ever be multithreaded.
Actually, it has been mentioned a lot of times that eve is hard locked into one thread, even server side, as a python limitation. I have not seen a Devblog saying that they solved it, so my guess is that this limitation still exists. Doesn't make me happy either, but ccp admitted that weakness multiple times over the years. |
Atherin Gaius
Caldari Domini Umbrus DEFI4NT
|
Posted - 2010.01.26 02:57:00 -
[39]
Originally by: Dr Ngo Also, they fixed Jita by giving it its own node and putting a cap on local. Nothing magical about it, but it would be unreasonably pricey to do for every system in eve.
A cap on local in Jita? It had 1500 people in it yesterday.
|
Van PokerAlho
Amarr Red Federation
|
Posted - 2010.01.26 03:55:00 -
[40]
Rumors inside CCP say that because of the super pandemic of H1N1 (is H1N1 right?, the A flu, allow me to rephrase it, the super wannabe deadly flu) that was going to strike the world in the winter, all the hamster running the game were vaccinated and most died due to vaccination problems.
Now I have two questions: Where is the deadly flu they promised us? Where is the so long promised global warming? I want warm, beach, nude girls and everything that comes with it!
|
|
Taedrin
Gallente The Green Cross DEFI4NT
|
Posted - 2010.01.26 04:08:00 -
[41]
Edited by: Taedrin on 26/01/2010 04:10:58
Originally by: Xiaodown
Originally by: Taedrin
Originally by: Garulf Hirgon Throw money at the problem: if they can't code for performance, buy better hardware. CCP should be swimming in cash...they're bringing in like $4 million USD a month and they can't have THAT much for overhead and operating expenses...
Because CPU clockspeeds are totally continuing to rise at an exponential rate.
OH wait... That's right. CPU clock speeds haven't increased since the pentium 4. Multiple cores? Don't make me laugh. Those can only increase performance in multi-threaded applications. EVE is single threaded. That means that each solar system in EVE can be handled by AT MOST 1 CPU core.
Sorry to burst your bubble there, but it is impossible to "buy better hardware", because better hardware just simply doesn't exist anymore. The fastest server CPU I can see on newegg is clocked at 3.3 GHz (costing 1,000 USD for the cheapest one). You will note that Tranquility's hardware is ALSO clocked at 3.3 GHz. CCP is already using the best hardware commercially available.
Dude, what?
Yes, all of eve is running on one single core 3.3Ghz proc.
There is NO WAY that serialized mathematical calculations could ever be multithreaded.
Are you seriously just spouting stuff you read on digg 5 years ago?
No, not all of EVE is running on a single core 3.3GHz processor. But each node is located on at most one 3.3GHz CPU core.
I'm also not pulling stuff out of thin air. CCP stated a few years back when they were talking about the infiniband stuff that they have no plans to multithread the solar system code (ie: nodes). They can put chat channels,
Consider this quote for example:
Originally by: Dev Blog The problem for us is the granularity of the load balancing. We have lots of very small grains, like a chat service for a solar system or is geographically abstracted and can therefore be located in another thread which physically translates to another CPU. Same goes for lots of bigger grains, such as market, corp services, alliance services etc.
However, we have on very big grain which we can today not split between threads without running the risk of handover problems, synchronization problems or even worse - players stuck in the middle of space.
Stuck usually happens only on session changes (docking, jumping) which is exactly when you are handed over between CPU's (or forced by a service which requires it). The big grain is the Solar System. Imagine being able to get stuck around a gate in combat. Not a pleasant thought :)"
source
So no. I'm not spouting stuff I read on digg 5 years ago. I'm spouting stuff I read from a devblog 5 years ago. ---------- There is always a choice. The choice might not be easy, nor simple, nor the options be what you desire - but, nevertheless, the choice is there to be made. |
Douglas Pearce
Caldari Music Martinis and Misanthropy
|
Posted - 2010.01.26 04:11:00 -
[42]
Originally by: Taedrin I'm spouting stuff I read from a devblog 5 years ago.
----------- I disagree |
Tiberizzle
Caldari Tactical Trading Partnership
|
Posted - 2010.01.26 04:28:00 -
[43]
Edited by: Tiberizzle on 26/01/2010 04:33:30 CCP didn't actually attempt to solve any of the problems involved in a single instance MMO. Instead, they called them game mechanics. Each starsystem is effectively a shard unto itself and every jump is effectively a character transfer between shards.
To the previous poster who implied that CCP did any sort of load balancing - bull****. There've been how many dev blogs / posts that've explained they run 1 starsystem per thread, and "unreinforced" nodes tend to run several starsystem threads per physical host. Moving a shard to a larger server during down time isn't load balancing in any environment I've worked in, its a ****ing hardware upgrade.
So they create this awesome game that LOOKS like a single instance MMO because you can character transfer between shards trivially, and chat/market/guilds are driven by the same data. The problem creeps in when a large number of the players in the single instance (on the order of the number of players a single shard in other MMOs can support -- how surprising) decide to congregate on the same shard, then the fact that they did not parallelize the load in any meaningful way rears its ugly head. Also, when they bury their head in the sand about the fact that their single instance MMO is smoke and mirrors, and try to design gameplay to encourage massive congregations of players.
They need to be able to parallelize the load of a single grid, and I strongly suspect its not a database issue. Rollbacks are a huge clue here. The system thread acts as a cache for objects in the system and I would guess primarily calculates ranges. Data is loaded from the DB, flushed on "session change", etc etc. When it dies due to load the cached data hasn't been flushed, hence the "rollback". Mutual range calculation is, most simply, O(n^2) for n objects. Range calculation is, obv., sqrt(dx^2+dy^2+dz^2). The sqrt can be avoided in many cases (e.g. when simply checking range and actual distance quantity is irrelevant), but where it cannot it makes for a particularly ugly O(n^2). There are much better algorithms that have some prayer of being parallelizable within the context of a cluster (specifically thinking a distributed spatial octtree here -- O(ln(n)), traversal and data store beyond a certain variable depth passed to subnode in the cluster based on load, you know how it goes), but from the behavior observed and past comments it doesn't seem to be the case that they're using any of them.
I've seen several multiplayer space-y games run into this problem, including one of my own. Every action requires a range check for everyone in the space to determine who observes it, and when you get a lot of ships in the same space that means (a lot)^2 of range checks. In the O(n^2) algorithm 1000 objects results in 10 million range checks for every action. Consider: drones, wrecks, cans, missiles for 1000 players. The ranges are probably cached when ships are moved, so every game tick could result in hundreds of millions of these calculations.
It's very possible that some simple optimizations will return <1500 players to playability, as it worked previously and seems to be just on the threshhold around that number anyway. It won't change the fact that their single instance MMO oh look at our cluster marketing spiel does not agree with the reality of their implementation, and it will continue to limit the number of objects that can be in play (see also in past patches: reduction of NPC ships in missions, reduction of drone control capacity, etc.) Another poster was astute in observing that IPC and per-core clock speed have not been moving much of late -- the problem won't really go away until they learn how to code the implementation they've been bull****vertising for the last 7 years.
[/words]
edit: lol, and python. What a joke. Stackless IO -- because one optimization trick will save us from the fact that the rest of the language is immature & slow as ****! Excellent.
|
SSDD24
Gallente tr0pa de elite Triumvirate.
|
Posted - 2010.01.26 04:51:00 -
[44]
CouchDB isnt that crazy and i think it would fit eve way better then the traditional DBs
|
Xiaodown
Caldari Guiding Hand Social Club Dystopia Alliance
|
Posted - 2010.01.26 05:30:00 -
[45]
Originally by: SSDD24 CouchDB isnt that crazy and i think it would fit eve way better then the traditional DBs
Can you and I fall in love?
Seriously, using couch for an internal project with the team that I work with now, and it's amazing. We love it.
But, then, we're using it as a directory of meta information, so..... YMMV.
Originally by: Taedrin
So no. I'm not spouting stuff I read on digg 5 years ago. I'm spouting stuff I read from a devblog 5 years ago.
Ugh, dude, that's painful. I guess I assumed that the process would go something like this: 1.) Think of ways to develop the game 2.) pick out the worst one; throw it away 3.) Pick one from the remaining ones.
Seems they went for 1.) think of ways to develop the game 2.) pick out the worst one; use it
I stand, shamefully, corrected. I hope that dev blog is out of date.
--
|
Taedrin
Gallente The Green Cross DEFI4NT
|
Posted - 2010.01.26 05:41:00 -
[46]
Edited by: Taedrin on 26/01/2010 05:42:23
Originally by: Xiaodown
Originally by: SSDD24 CouchDB isnt that crazy and i think it would fit eve way better then the traditional DBs
Can you and I fall in love?
Seriously, using couch for an internal project with the team that I work with now, and it's amazing. We love it.
But, then, we're using it as a directory of meta information, so..... YMMV.
Originally by: Taedrin
So no. I'm not spouting stuff I read on digg 5 years ago. I'm spouting stuff I read from a devblog 5 years ago.
Ugh, dude, that's painful. I guess I assumed that the process would go something like this: 1.) Think of ways to develop the game 2.) pick out the worst one; throw it away 3.) Pick one from the remaining ones.
Seems they went for 1.) think of ways to develop the game 2.) pick out the worst one; use it
I stand, shamefully, corrected. I hope that dev blog is out of date.
I do too, actually. I was one of the voices back then calling for multi-threading the server code to allow for thousands upon thousands of players in the same solar system. Was very disappointing when they told us that they couldn't/wouldn't. A man can dream though...
I am personally of the opinion that they should have worked on updating the code, even if they had to take a significant portion of the dev team to the task for a year or two. But they are a business and probably don't want to spend that much money on the issue. ---------- There is always a choice. The choice might not be easy, nor simple, nor the options be what you desire - but, nevertheless, the choice is there to be made. |
hepatitisDD
Caldari GoonFleet GoonSwarm
|
Posted - 2010.01.26 06:21:00 -
[47]
The obvious solution is to drop sov all over the universe, thereby removing any incentive to have large fleet fights. Amirite?
|
Hinkledolph
Minmatar Viper Squad Triumvirate.
|
Posted - 2010.01.26 07:13:00 -
[48]
Originally by: bloody johnroberts you just have to watch the great fanfest vid on dominion and listen to there great statements on how it would change the way eve plays lol
yep we see it nice planets lol but game is unplayerble
everyone knows its the patch that has caused this but ccp wont agree because its a bad thing to say we screwed up as they are using dominion as a selling tool lol sandbox more like ashtray
Dragon fleet is scary
|
Ed Rush
Gallente Erasers inc. Mostly Harmless
|
Posted - 2010.01.26 09:43:00 -
[49]
world of warcraft has some kind of capture the flag that handles 100 or so.. maybe they have even more?
their problem is for what i've read, is that their solar system node code cannot span onto multiple physical processing units. and they'd need to rewrite the code... and you know.. that is not going to happen because it costs REAL LIFE ISK :D
|
Niina
Minmatar Eve Defence Force Systematic-Chaos
|
Posted - 2010.01.26 12:10:00 -
[50]
Originally by: Ed Rush world of warcraft has some kind of capture the flag that handles 100 or so.. maybe they have even more?
their problem is for what i've read, is that their solar system node code cannot span onto multiple physical processing units. and they'd need to rewrite the code... and you know.. that is not going to happen because it costs REAL LIFE ISK :D
Actually, its not gonna happen due the fact that all transfers from CPU to another have high possibility of getting out of synch, resulting in even more moan. What they could do is remove drones and missiles.
Game mechanic wise its a terrible loss ( i love my gallente boats :( ) but every object with numeric data that changes = more calculations, meaning less ppl on same node.
And if we go as far as rollback, please CCP, roll back to RMR ^^.
|
|
Spike Spegel
Minmatar Black Viper Nomads
|
Posted - 2010.01.26 13:52:00 -
[51]
http://www.howstuffworks.com/dna-computer.htm
Spike out |
ElfeGER
Caldari Versatech Co. Aeternus.
|
Posted - 2010.01.26 17:43:00 -
[52]
The technology to fix lag is available today. Look at Google for example they scale and scale.
CCP is also doing something to improve the situation. Drones ROF was changed, weapon grouping introduced (are there any internal stats if people stopped using that because of bugs with stuck modules - one fc that has his fleet with ungrouped mods can nuke a lot of cpu cycles)
On the hardware side there is not much to improve on the cluster as the cpu are already as fast as possible per core and the db load/lag should be ok (or else the api would be off :D )
The main issues are in the Eve code. - no dynamic load balancing - single core for a sol node
Performance profiling can do wonders to code when you can redo the big cpu hogs. The issue is just to identify those first which needs testing or better automated testing of those large scale events. So the server does less work and delegates work where ever possible to different nodes.
|
Morvyn
Caldari Destry's Lounge Important Internet Spaceship League
|
Posted - 2010.01.26 17:45:00 -
[53]
Posting ITT to confirm that MSSQL is the best database, Python is the best choice of language for server code, and that clusters that do load balancing are lame.
Originally by: SkeDOOSH They are godless killing machines that put even grizzly bears to shame.
|
Kythren
Caldari Jelly Baby Corporation Majesta Empire
|
Posted - 2010.01.26 18:01:00 -
[54]
i heard python is the solution, oh wait...
fyi the reason why ccp employs pythons instead of strong irish working hamsters is that they want to deploy several expansions per year. however they actually fail to understand that quantity of expansions isnt better than quality of code when it comes to mmorpg design. since mmorpg means A LOT of clients which transmit and receive data and the complexity of calculations in e.g. eve rise with the client count exponentially, having excellent code is a must have! ;) ----- no u! support my CSM petition <sarcasm="inside">
|
Shander Maxum
Gallente The 5th Freedom
|
Posted - 2010.01.26 19:23:00 -
[55]
It doesn't need to be all or nothing..
I think thats a major leap that would need to be made...trying some new things doesn't need to mean immediately throwing out the old.
Certain new 0.0 regions(wormholes?) could be added or some in existing regions could be changed that operate under different rules that might lend themselves to different players time availability, alliance sizes etc..
Some 0.0 systems might depend more on player activities to hold ... some might easier to defend than others, or maybe you'd get concord defence of sovereignty unit (but not of player ships) in one system per 50 accounts if you killed enough rats monthly etc.. giving entry to 0.0 to another sort of group....But NOT at the expnse of the current system. (things different than low sec and different than current normal 0.0 )
Or like you brought up, maybe you'd have some systems with limits on number of hostile pilots allowd in them (but would you limit blues too, if it were 100 vs 100 would that mean 100 lower spell point newer players would be overwhelmed with gangs of 100 titans without being able to call for any support from friends ?)
Using the worm-hole system might be a start maybe with the ability for a corp to earn one permanent hole to somewhere by fulfilling certain requirements (and maybe with some limitted control to determine where that whole would lead to and abilty to move it at a decent cost and limited frequency) with others limited by what they could jump in (perhaps with larger deterioration thresholds or rules)
There are probably better brainstorms but
I think the first major leap would be that not all Systems need to be the same and that by experimenting with a sort of parallel system players would vote with their feet, gravitating to what they might find more interesting .
|
Tiberizzle
Caldari Tactical Trading Partnership
|
Posted - 2010.01.27 20:20:00 -
[56]
lol @ spell point
I wanna cast small energy turret!
|
Reno Shinra
Caldari Diversity 101
|
Posted - 2010.01.27 20:29:00 -
[57]
why not turn eve into wow and have differant servers for pvp and 1 for mission runners 1 for 00 and 1 for carebears - no its doesnt work - strange the last 5 years have been patch after patch more content - server get slower and slower - maby its time to upgrade ya server instead of running it on an old vic20 and spend some of ya budget on making agame that playable.
Oh and Cva free the slaves !
|
Cupdeez
Caldari Out of Order Cult of War
|
Posted - 2010.01.27 22:09:00 -
[58]
Its not the amount of hit points because if people can make a gang that size then why not bring a gang that size..
I think CCP should start thinking about ways and rules of systems...
Example:
If the system has 50 or more dreads in siege all the incomfing gates in the system lock down. (Jumping into _____ is currently offline because of an abnormal fluxation in space time is preventing it)
When this happens the only way into the system is using a cyno and the only ship able to make it through the abnormal fluxation are capitals..
If other ships try and jump through a titan cyno bridge 80-90% of them will be crushed so and destroyed because of the fluxaction.
hehehe just an idea
possibly have a speical ship class that can still jump through the gate to get cyno's into the system.. "kinda of a non-pvp ship so you don't see 200 man gangs of these haah.
Out side the BOX.
|
Depili
Caldari Blood Works Inc. Circle-Of-Two
|
Posted - 2010.01.27 22:58:00 -
[59]
I don't mind little lag in the larger battles, as long as it's just lag.
Currently we get all kinds of anomalies, stuck guns, weird warps, ghost ships, ships exploding when they have warped out etc, they should be eliminated. Also lag should affect everyone the same.
A fleet fight with 1min lag would be bearable if it would affect each module the same, they would still cycle (even if they would all cycle the same % slower) as long as they would still autorepeat and not get stuck.
Now we get a dice throw every time with lag related anomalies and how the lag affects your and your enemies fleets.
|
Anela Cistine
Amarr GoonFleet GoonSwarm
|
Posted - 2010.01.27 23:42:00 -
[60]
Those anomalies are great. You can see ships warp into the shields of a safe POS and then explode from damage they took 5 minutes ago. That is actually pretty handy, since it lets you scoop the loot from your ship at your leisure. The day after a big buggy battle I noticed 3 empty pods sitting inside the POS shields, I don't know WTF happened there. Once I was in my ship which appeared to be on fire but safely burning away, until I noticed that I could see my own corpse sitting next to my burning ship. Oh, I guess I must be dead. I kind of wish I hung out with roleplayers in EVE. I'd love to see how they rationalize that stuff in character.
The worst is when you jump into a system and get trapped in a session change, you can't see and you can't move (though you can still be seen and be shot at) and worst of all you can't even log out. Not being able to log out is inexcusable. Being locked in a game blind and paralyzed isn't any one's idea of fun. You can't even file a stuck petition, because pushing escape doesn't take you to the options screen, it just locks you into a black screen forever.
|
|
|
|
|
Pages: 1 [2] 3 :: one page |
First page | Previous page | Next page | Last page |