EVE Search - What if there is no good solution to the node overlaod probelm?

All Channels

Corporation, Alliance and Organization Discussions

What if there is no good solution to the node overlaod probelm?

» Click here to find additional results for this topic using Google

Monitor this thread via RSS [?]

Pages: [1] 2 3 :: one page
Author	Thread Statistics \| Show CCP posts - 0 post(s)
Lydai Caldari Inner Light Creations	Posted - 2010.01.25 18:03:00 - [1] Please correct me, but I know of no MMORPGs, which can handle even 100 clients/player characters in the same zone without serious lag much less 600+. 1. What if it is the case, that with current processors and network technology, their is no real fix for this problem? 2. Doesn't the (both current and past) magnitude of lag indicate that no simple tweak or additional hardware is going to solve this problem. Is reducing grid loading time from 30 minutes to 30 seconds a fix, or going from 1/fps to 5/fps a a solution. Instead of demanding some magical tech fix or programing breakthrough, which will allow large fleet battles to take place in some acceptable orderly fashion, people may need to consider that this is not in the cards; not now, maybe not in the future. People may need to face the fact that "bring all ya got" fleet battles will not be a feature of this game, ever. Minor or even moderate fixes to the 0.0 sovereignty system will not work(and in the case of Domain could make things worse). I might be,that what is required is a major, dramatic, game altering, ground up, total change to the Sovereignty system. One which, while by necessity limit the number of simultaneous clients/player characters in a zone and thus; one which, while also by necessity, will anger, infuriate and alienate a large part of the 0.0 player base.
Garulf Hirgon Gallente Burning Technologies Circle-Of-Two	Posted - 2010.01.25 18:07:00 - [2] 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... Inappropriate Signature Locked. StevieSG
Ga'len Amarr Minmatar Ship Construction Services Ushra'Khan	Posted - 2010.01.25 18:09:00 - [3] Edited by: Ga''len on 25/01/2010 18:11:11 Originally by: Lydai Please correct me, but I know of no MMORPGs, which can handle even 100 clients/player characters in the same zone without serious lag much less 600+. 1. What if it is the case, that with current processors and network technology, their is no real fix for this problem? 2. Doesn't the (both current and past) magnitude of lag indicate that no simple tweak or additional hardware is going to solve this problem. Is reducing grid loading time from 30 minutes to 30 seconds a fix, or going from 1/fps to 5/fps a a solution. Instead of demanding some magical tech fix or programing breakthrough, which will allow large fleet battles to take place in some acceptable orderly fashion, people may need to consider that this is not in the cards; not now, maybe not in the future. People may need to face the fact that "bring all ya got" fleet battles will not be a feature of this game, ever. Minor or even moderate fixes to the 0.0 sovereignty system will not work(and in the case of Domain could make things worse). I might be,that what is required is a major, dramatic, game altering, ground up, total change to the Sovereignty system. One which, while by necessity limit the number of simultaneous clients/player characters in a zone and thus; one which, while also by necessity, will anger, infuriate and alienate a large part of the 0.0 player base. So....what you are saying is that once they have rolled out an expansion that has broken the game they should go about and completely change the game into another incarnation that vaguely resembles the old game. Hmm...where have we seen that before. Star Wars Galaxies, releases, Combat Upgrade, then releases Combat Upgrade II, then releases NGE. Yeah, that worked out well for them, didn't it? No, a rollback of the entire expansion may be the only option.
Scatim Helicon Caldari GoonFleet GoonSwarm	Posted - 2010.01.25 18:12:00 - [4] We've seen that the massive fleet fights are possible and playable. It was possible 2 months ago, but Dominion broke something. -----------------
Swren1 Gallente Infinite Covenant Looney Toons.	Posted - 2010.01.25 18:35:00 - [5] Edited by: Swren1 on 25/01/2010 18:35:56 nvm i'm drunk
Pixel SonursCreen Gallente Volatile Nature Systematic-Chaos	Posted - 2010.01.25 18:42:00 - [6] they were able to fix the 1000+ in jita about a year ago also, bring all ya got massive fleet battles are what is eve about. suggesting to fall back on "smaller fleets" will just mean u have no chance to reinforce anything, not to mention conquer space of your enemies. this would mean 0.0 would turn into a high sec warzone with bubbles. eventually solidifying in an unattractive situation where all is decided and nothing can be changed. if that isnt allready yet the case with the current situation after dominion hit the shelves. "oh lets spread out abit, to avoid lag,... 18 dreads here, 12 dreads there, camp that gate with 15 and camp the station with 23 guys 4 jumps away!" "oh yeah, good idea! cool pvp" "oh ****, our targets dont like us outhere, they just brought 120 battleship fleet to get us out of here" what you are saying makes no sense. someone clearly made a mistake with the dominion expansion. the timers and hitpoints of the new sov structures do not suggest 0.0 alliances should fall back on smaller fleet warfare what so ever.
Shanky McStabber Minmatar Merch Industrial GoonSwarm	Posted - 2010.01.25 19:31:00 - [7] Originally by: Scatim Helicon We've seen that the massive fleet fights are possible and playable. It was possible 2 months ago, but Dominion broke something. THIS. The problem really stems from the fact that 2 months ago you could bridge people into a system without the node exploding. You could also have massive fleet fights without node death. Then Dominion comes out and the node goes haywire with just 400 people in system.
Red Armageddon Caldari CASCADE OF SPECTRES Emergence.	Posted - 2010.01.25 20:12:00 - [8] Originally by: Pixel SonursCreen they were able to fix the 1000+ in jita about a year ago But what did they do to fix it? Oh its just Red Armageddon don't worry about him, hes harmless..... I swear I'm harmless. :D
Meno Theaetetus Gallente Wildly Inappropriate Wildly Inappropriate.	Posted - 2010.01.25 20:14:00 - [9] Edited by: Meno Theaetetus on 25/01/2010 20:15:05 Perhaps time to move away from off the shelf solutions, such as ms sql, etc. Time to move mmo's onto proper clusters that distribute load over the whole cluster, distributing work over the cluster based on some kind of node per system/region, is far from optimal. In fact calling the tq a 'cluster' is a little like calling a group of pcs in a room all accessing the internet(shared resource) a 'cluster'. Clusters are about dynamic load balancing, something that tq does not do.
Mya ElleTerego Amarr The Hull Miners Union Gentlemen's Club	Posted - 2010.01.25 20:18:00 - [10] Edited by: Mya ElleTerego on 25/01/2010 20:21:04 Being realistic, if they dont fix it, I forsee alot of turtling for dominion sov warfare. With people generally giving up once they get to main systems of coalitions. That will protect thier main assets and keep them from being homeless, but they will have thier work cut out for them as their empires burn around them. Good thing is quite a few alliances have some spare pos laying around, post dominion as replacements. Also making sbu, ihub, tcu, and stations have obscene amts of hitpoints, doesnt help mitigate blob warfare. Im glad CCP took the time last patch to add a nice buff to thier hp as if 200 dudes shooting them for 30 min wasnt enough. Alliance Recruit thread Alliance Homepage/Killboard
Gamine Amarr Rage For Order Systematic-Chaos	Posted - 2010.01.25 20:25:00 - [11] Originally by: Scatim Helicon We've seen that the massive fleet fights are possible and playable. It was possible 2 months ago, but Dominion broke something. As much as I hate to agree with a Goon.....
kyrieee Gallente Psykotic Meat Wildly Inappropriate.	Posted - 2010.01.25 20:28:00 - [12] I think EVE's codebase it too old and too arcane to be salvageable, even though that's just pure speculation on my part. I mean, the features they're able to add are, even though they sometimes have a strong effect on gameplay, very minimal and tend to be modifications of pre-existing systems in the game. At the same time you always hear about how they can't do this or that because of the code and how they don't even have the guys who wrote it on staff anymore and they need to spend a year just trying to understand it. Yes, EVE has been and still is going strong but if they want to continue that way for 3-4 more years I think they'll have to add significant new features to the game and it seems they're not able to. A complete re-write of the code base isn't really a realistic undertaking though, but I could see CCP going for it
Xiaodown Caldari Guiding Hand Social Club Dystopia Alliance	Posted - 2010.01.25 20:32:00 - [13] Originally by: Meno Theaetetus Edited by: Meno Theaetetus on 25/01/2010 20:15:05 Perhaps time to move away from off the shelf solutions, such as ms sql, etc. Time to move mmo's onto proper clusters that distribute load over the whole cluster, distributing work over the cluster based on some kind of node per system/region, is far from optimal. In fact calling the tq a 'cluster' is a little like calling a group of pcs in a room all accessing the internet(shared resource) a 'cluster'. Clusters are about dynamic load balancing, something that tq does not do. Take it from a cluster admin, it's a hard problem to solve. TQ does dynamic load balancing (in the meaningful sense), but your bottleneck will always be the one central authority for database information. It's relatively easy (almost trivial to those of us who do it for a living) to create any number of read-only database slaves and internally load balance them. It is significantly harder to handle lots and lots of writes. Add in all the event notifications and message passing, and woah. I mean, how do you handle a brazillian writes? Well, you get a DAS or a SAN, and you fill the thing up with 90 15krpm 2.5" SAS drives, and make it one giant raid10. When that doesn't have the I/O, what then? Well, you get a RAMSAN, which is the same thing, but remove the hard drives and replace with RAM. CCP has already done that, where do you go now? I honestly don't know. I guess you start seeing how heavily you can shard the information. Anyway, you have some odd ideas about clustering - clustering for high availability and load balancing is not a panacea for load issues. You are thinking of beowulf clusters, something that's essentially unheard of outside of the scientific and financial communities. That's the "we have one question and many computers to compute part of the answer" cluster. The cluster that TQ and most other solutions do is "we have many questions and many computers to answer them". Traditionally, these are handled with solutions like BigIP F5's, Linux LVS/KeepaliveD, Whack-a-mole style IP sharing, etc. The problem is that you can't always just throw more hardware at it - at some point, there's persistent connections and persistent information that must be maintained. I don't have the answers, but it's not a plug-and-play-more-hardware fix. No, the biggest concerns I have are ENTIRELY software based - I don't feel that this is an architecture issue. There is no way to use load balancing to explain the fact that 1.) It was playable 2.) Patch 3.) It is now not playable That's not hardware. As for Quote: Perhaps time to move away from off the shelf solutions, such as ms sql, etc. I think you may be right. Might be time to look at some really wacky solutions, like Erlang/CouchDB or some other crazy key-value-store system that may be faster. Or extreme architecture changes like excessive sharding of data storage. Who knows? --
Abriana Overlord Caldari Evolution IT Alliance	Posted - 2010.01.25 20:38:00 - [14] Originally by: Meno Theaetetus Edited by: Meno Theaetetus on 25/01/2010 20:15:05 Perhaps time to move away from off the shelf solutions, such as ms sql, etc. Time to move mmo's onto proper clusters that distribute load over the whole cluster, distributing work over the cluster based on some kind of node per system/region, is far from optimal. In fact calling the tq a 'cluster' is a little like calling a group of pcs in a room all accessing the internet(shared resource) a 'cluster'. Clusters are about dynamic load balancing, something that tq does not do. Yeah start with oracle look at most real heavy duty database out there and they will 9/10 time pick oracle over SQL There is bound to be a level dynamic load balancing but it also depends on how that load split and at what layer. My guess is that Jita has its own dedicated hardware to an extent when it comes to large fleet type battle the scale of that load is too much for a server so comes down tohow the hardware is distrubted in terms of horsepower. I do wonder where the bottleneck is though would be interesting to see the output from the upcomming stress testing is the load hitting cpu, memory, storage etc and how that ripples through the tier (apps, database, presentation etc) I am not overly familiar with the load balancing they are using I am guessing it is a mix between system and hardware load balancings, F5 are awsome but can be troublesome also. Software loadbalaning only takes things so far. I am guessing the overall lagmonster is a combination between some "less than optimal coding" coupled with infrastructural concerns. No motter how it goes though, what CCP has accomplished is very good technologically
Red Armageddon Caldari CASCADE OF SPECTRES Emergence.	Posted - 2010.01.25 20:49:00 - [15] I honestly think its time for CCP to think outside the box. Either, Work very very hard, At that issue. Or really take a look at how to scale down pvp. Because adding massive hitpoints to a structure, does not help small fleet warfare. Only promotes big stuff. Idk what they are planing. But i think the player base needs some encouragement, Atleast a small morale boost. Oh its just Red Armageddon don't worry about him, hes harmless..... I swear I'm harmless. :D
Meno Theaetetus Gallente Wildly Inappropriate Wildly Inappropriate.	Posted - 2010.01.25 20:52:00 - [16] Without a doubt its very difficult problem, and as you say it comes from massive amounts of writes, with those changes needing to be reflected back to the client at the same time. When I was talking about trying to balance the load I was referring to the fact that in the current model they use, they try and represent a grid or a system on one node, whilst the rest of the cluster is sat there doing nothing. With a theoretical hat on of course, why not try a model where players locations, ship locations, etc are distributed over the cluster and there location in the database are not representative of there geographic location in eve. It would require an added abstraction layer between the node db servers and the backend logic, perhaps with some kind of sql multiplexers to cross reference it. But then you may just be transferring the lag onto the multplexers and will not see any performace gain. It would be interesting to see how much lag is generated based on database I/O and how much lag is generated from number crunching the physics. An obvious answer to this would be some kind of result forwarding, like you see in processing pipelines, however this is where the 'desynch' would happen. Perhaps the recent performance problems have stemmed from the work they put in to fix the desynch problems.
CryptorsChild Caldari Trojan Trolls Mostly Harmless	Posted - 2010.01.25 21:13:00 - [17] I would like an in-depth tour of the EVE datacenter and how it's managed. That would be awesome for nerds. :D
General Windypops Gallente The Littlest Hobos Ushra'Khan	Posted - 2010.01.25 21:26:00 - [18] The main problem eve faces in a high load situation is basically that there's packet loss between the 2 key parts of the client and server code, one that picks up the 'tell' or 'initiator' and the receptor that turns that into an action on your screen. At any one moment in time there might be 10s of thousands of these little data binds taking place (particularly when people are doing server intensive activites like activating a pulse laser, or typing into a busy chat channel) as this forces the database to query any number of possible outcomes and then use those scenarios to decide the most likely of them (this is called Stochastic data transfer). Extrapolate this over 1,000 people in a system and you'll see why lag on this scale is pretty much unavoidable. One of the biggest problems that CCP probably have is the geographic distance between the 'sending' and their 'receiving' servers. One is in Iceland (which is the server that decides what you should see on the screen) and the other at their datacenter which I believe is in London (the server that does the calculation). Twitter:@genrlwindypops
Rani McGee Minmatar The Resurging Wind Lost Sheep Domain	Posted - 2010.01.25 21:32:00 - [19] I wonder if they've looked into using Tesla chips which can handle thousands of threads on just one chip. Seems like a highly parallel problem which those chips could solve. And they aren't even that expensive (well maybe to you and me but not for a company with some money).
Dr Ngo Amarr Shut Up And Play WE FORM VOLTRON	Posted - 2010.01.25 21:34:00 - [20] Better hardware would be nice, but I believe the current problems stem mostly from the recent changes to the way fleets work. If they have to scrap a new system to make the game playable it'll probably end up being the domi fleet system. 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.
Lydai Caldari Inner Light Creations	Posted - 2010.01.25 21:34:00 - [21] "We've seen that the massive fleet fights are possible and playable. It was possible 2 months ago, but Dominion broke something." What do you mean by "massive"? 300? 600? 1200? What do you mean by "playable" ? 30? 60? 300? secs to load grid or activate mod [Just asking, Not disagreeing, I don't have a lot of direct knowledge, ] if CCP was able to fix things so that the lag was no worse than before Domain and maybe over the next year improved that 25%, would you find that acceptable and good enough?
YChin Mei Caldari The All-Seeing Eye Atlas Alliance	Posted - 2010.01.25 21:35:00 - [22] Edited by: YChin Mei on 25/01/2010 21:37:23 Originally by: General Windypops The main problem eve faces in a high load situation is basically that there's packet loss between the 2 key parts of the client and server code, one that picks up the 'tell' or 'initiator' and the receptor that turns that into an action on your screen. At any one moment in time there might be 10s of thousands of these little data binds taking place (particularly when people are doing server intensive activites like activating a pulse laser, or typing into a busy chat channel) as this forces the database to query any number of possible outcomes and then use those scenarios to decide the most likely of them (this is called Stochastic data transfer). Extrapolate this over 1,000 people in a system and you'll see why lag on this scale is pretty much unavoidable. One of the biggest problems that CCP probably have is the geographic distance between the 'sending' and their 'receiving' servers. One is in Iceland (which is the server that decides what you should see on the screen) and the other at their datacenter which I believe is in London (the server that does the calculation). I am not sure I understand you Windy...Are you suggesting that we should pull Iceland closer to England? If that is the case, I am in! Grab the hooks guys. I'll do everything for the greater good
General Windypops Gallente The Littlest Hobos Ushra'Khan	Posted - 2010.01.25 21:39:00 - [23] Originally by: YChin Mei Originally by: General Windypops The main problem eve faces in a high load situation is basically that there's packet loss between the 2 key parts of the client and server code, one that picks up the 'tell' or 'initiator' and the receptor that turns that into an action on your screen. At any one moment in time there might be 10s of thousands of these little data binds taking place (particularly when people are doing server intensive activites like activating a pulse laser, or typing into a busy chat channel) as this forces the database to query any number of possible outcomes and then use those scenarios to decide the most likely of them (this is called Stochastic data transfer). Extrapolate this over 1,000 people in a system and you'll see why lag on this scale is pretty much unavoidable. One of the biggest problems that CCP probably have is the geographic distance between the 'sending' and their 'receiving' servers. One is in Iceland (which is the server that decides what you should see on the screen) and the other at their datacenter which I believe is in London (the server that does the calculation). I am not sure I understand you Windy...Are you suggesting that we should pull Iceland closer to England? If that is the case, I am in! Grab the hooks guys. I'll do anything for the greater good Just trying to dispell some of the nonsense that people are posting around here. Quite a few of my friends work in IT departments and I also once built an access database for a summer job. I have done my bit trying to pull Icelandic girls, on the off chance that this, too, might assist us. Twitter:@genrlwindypops
Dr Ngo Amarr Shut Up And Play WE FORM VOLTRON	Posted - 2010.01.25 21:50:00 - [24] Originally by: YChin Mei Edited by: YChin Mei on 25/01/2010 21:37:23 Originally by: General Windypops The main problem eve faces in a high load situation is basically that there's packet loss between the 2 key parts of the client and server code, one that picks up the 'tell' or 'initiator' and the receptor that turns that into an action on your screen. At any one moment in time there might be 10s of thousands of these little data binds taking place (particularly when people are doing server intensive activites like activating a pulse laser, or typing into a busy chat channel) as this forces the database to query any number of possible outcomes and then use those scenarios to decide the most likely of them (this is called Stochastic data transfer). Extrapolate this over 1,000 people in a system and you'll see why lag on this scale is pretty much unavoidable. One of the biggest problems that CCP probably have is the geographic distance between the 'sending' and their 'receiving' servers. One is in Iceland (which is the server that decides what you should see on the screen) and the other at their datacenter which I believe is in London (the server that does the calculation). I am not sure I understand you Windy...Are you suggesting that we should pull Iceland closer to England? If that is the case, I am in! Grab the hooks guys. I'll do everything for the greater good Typical, making everything harder than it has to be. It'll be much easier to pull England closer to Iceland.
YChin Mei Caldari The All-Seeing Eye Atlas Alliance	Posted - 2010.01.25 21:54:00 - [25] Originally by: General Windypops Just trying to dispell some of the nonsense that people are posting around here. Quite a few of my friends work in IT departments and I also once built an access database for a summer job. I have done my bit trying to pull Icelandic girls, on the off chance that this, too, might assist us. Well, personally I've never got past FoxPro tbh, DB stuff is just not in my area of interest. I hope CCP read the things you typed in here though, who knows there's no such thing as 'a useless idea'. Anything can help sometimes. You got me with your note about the Icelandic girls however. If this is a sacrifice we have to make in order to improve the situation, then...I stand ready. And..erhm, my bad, Dr Ngo.
Scatim Helicon Caldari GoonFleet GoonSwarm	Posted - 2010.01.25 21:58:00 - [26] Originally by: Lydai Quote: "We've seen that the massive fleet fights are possible and playable. It was possible 2 months ago, but Dominion broke something." What do you mean by "massive"? 300? 600? 1200? What do you mean by "playable" ? 30? 60? 300? secs to load grid or activate mod [Just asking, Not disagreeing, I don't have a lot of direct knowledge, ] During Delve II we had 2000 people in J-L fighting (so I'm told, I missed that one as I was busy evacuating assets from the east). Laggy, delayed module activation and sticky guns and similar glitches, but playable. I've been in multiple fights over the last 12 months where local was 1000 or more. Laggy, but playable. Even on occasions where there wasn't the opportunity to get a reinforced node arranged (1200 people in C9N around the end of 2008) the bugs and glitches were crazy but the system was semi-playable even if we spent as much time fighting the system as the enemy fleet. In Dominion, the system completely breaks - in the Y-2 "fight" half our fleet jumped in and loaded moderately quickly, the other half were still stuck in jump limbo 4 hours later. Quote: if CCP was able to fix things so that the lag was no worse than before Domain and maybe over the next year improved that 25%, would you find that acceptable and good enough? Dude, give back the server performance we had pre-Dominion and everyone in 0.0 warfare will be jumping up and down in joy. -----------------
General Windypops Gallente The Littlest Hobos Ushra'Khan	Posted - 2010.01.25 22:00:00 - [27] 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. Twitter:@genrlwindypops
Dark Sidhe Minmatar Rage For Order Systematic-Chaos	Posted - 2010.01.25 22:02:00 - [28] Given the history of the current fleet problems, it likely software based. We all know the tracking down coding errors is a pain but the magnitude of the issue is such they should focus on nothing else as far as new content. As it is, a sizeable chunk of the game is unplayable. ------------------------------- "Back away from the Tatoo Ink"
Scatim Helicon Caldari GoonFleet GoonSwarm	Posted - 2010.01.25 22:14:00 - [29] Originally by: Gamine Originally by: Scatim Helicon We've seen that the massive fleet fights are possible and playable. It was possible 2 months ago, but Dominion broke something. As much as I hate to agree with a Goon..... Its cool man, as demonstrated by the various Dev blogs before Dominion our real enemy is CCP. -----------------
gala gidas Amarr Parental Control Cult of War	Posted - 2010.01.25 22:48:00 - [30] 1)place the SBUs at the gates but not in the same system,but at the gates in the next system. 2)Link those SBUs so they have some kind of "common" HPs. 3)Create some "black hole effect" around the SBUs that limit the numbers-size of ships entering the area-grid. 4)Make possible for players to progress to the next level(hubs or stations that need to be reinforced)to use only the passage through those SBUs that are conquered,only to lead them to a new area that again is limited from a "black hole effect". 5)Allow warp outs in those areas but no warp-ins if the areas are already "full". Yes i dont know how many months of work this would take,but i can see that it clearly cuts huge fleets that wont be able to fight anyway to smaller and playable fights.If some systems are dead ends the surround systems can be used so multiple SBUs can be used. If this is doable the alliances that want to bring 1500 chars to a fight will still be able to do it but they will also need to find multiple FCs etc.

Pages: [1] 2 3 :: one page
First page \| Previous page \| Next page \| Last page

COPYRIGHT NOTICE
EVE Online, the EVE logo, EVE and all associated logos and designs are the intellectual property of CCP hf. All artwork, screenshots, characters, vehicles, storylines, world facts or other recognizable features of the intellectual property relating to these trademarks are likewise the intellectual property of CCP hf. EVE Online and the EVE logo are the registered trademarks of CCP hf. All rights are reserved worldwide. All other trademarks are the property of their respective owners. CCP hf. has granted permission to EVE-Search.com to use EVE Online and all associated logos and designs for promotional and information purposes on its website but does not endorse, and is not in any way affiliated with, EVE-Search.com. CCP is in no way responsible for the content on or functioning of this website, nor can it be liable for any damage arising from the use of this website.