EVE Search - Does anyone really believe something can be done about lag?

All Channels

EVE General Discussion

Does anyone really believe something can be done about lag?

» 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)
Hieronimus Rex	Posted - 2008.07.26 02:16:00 - [1] Edited by: Hieronimus Rex on 26/07/2008 02:16:44 Other than simply pack more people into the system before it lags.... For example... Right now Jita has its own node and can pack 700 or so people in with a fair bit of lag. This means that when there are 700 people the 701st person is contemplating entering Jita but chooses not to because of lag concerns (often I am such a person on Sunday afternoons). If 700 people could be in Jita on Sunday with no lag, there would probably be more people in Jita, say 1500 (myself included), until it lagged again. Same idea with fleet battles... the 201st person considering going on a fleet op might choose not to because of lag worries. If CCP attempts to combat lag blobs will only get bigger. Of course plenty of "ohlol it's lagging in Motsu, Jita, and fleet battles" threads are simply trolls or idoits, but presumably there is a nonzero number of threads posted by people who really actually believe CCP can do something about lag. I'm looking for their input.
Tarminic 24th Imperial Crusade	Posted - 2008.07.26 02:18:00 - [2] Well, a truly dynamic system would be able to divide combat/collision calculations across nodes in the entire Tranquility cluster, but allowing multiple nodes to handle combat on a single grid is a massive technical hurdle (from what the devs have said). Eventually? Yes. Soon? No. ---------------- Play EVE: Downtime Madness v0.83 (Updated 7/3)
Hieronimus Rex	Posted - 2008.07.26 02:19:00 - [3] Originally by: Tarminic Well, a truly dynamic system would be able to divide combat/collision calculations across nodes in the entire Tranquility cluster, but allowing multiple nodes to handle combat on a single grid is a massive technical hurdle (from what the devs have said). Eventually? Yes. Soon? No. Are we talking massive like rewriting EVE?
Tarminic 24th Imperial Crusade	Posted - 2008.07.26 02:25:00 - [4] Originally by: Hieronimus Rex Originally by: Tarminic Well, a truly dynamic system would be able to divide combat/collision calculations across nodes in the entire Tranquility cluster, but allowing multiple nodes to handle combat on a single grid is a massive technical hurdle (from what the devs have said). Eventually? Yes. Soon? No. Are we talking massive like rewriting EVE? Pretty much - the server-side combat engine would have to be rewritten to be multithreaded, which is both a design hurdle in itself (multithreaded code done right ain't easy) and if I remember correctly, the limitations of using Stackless Python also play a part, though I'm not familiar enough with the language to say exactly what that limitation is. I seem to remember the devs saying that at one point they were rewriting the most computationally intense parts of the server software in C/C++, but I can't remember where I heard that. ---------------- Play EVE: Downtime Madness v0.83 (Updated 7/3)
Sovereign533 Caldari PPN United Against ALL Authorities	Posted - 2008.07.26 02:29:00 - [5] Originally by: Hieronimus Rex Originally by: Tarminic Well, a truly dynamic system would be able to divide combat/collision calculations across nodes in the entire Tranquility cluster, but allowing multiple nodes to handle combat on a single grid is a massive technical hurdle (from what the devs have said). Eventually? Yes. Soon? No. Are we talking massive like rewriting EVE? yes, probaly the things needed for that (and i'm not going into detailles, and i'm gonna miss a LOT of them) are from the top of my head : multi-core support for the eve program, the program language either needs to be changed, or updated. good and fast communications between different nodes. good syncronisation between the nodes. make backup systems in case of an a-synchronisation (nodes shouldn't go conflict eachother, right?). multiple servers to accept and load the connection to the client. now everybody in a system is connected to 1 single blade server, this also needs to split up to get more bandwith. and this is just really basic. CCP is working with IBM and Microsoft to build a new supercomputer to reduce the lag even further. when it's done it'll be the fastest supercomputer used for gaming. if it will be enough, don't know. if it'll ever completely eliminate lag, uhm, no. the problem with lag is, it's just increasing the bar when it lags. lets say it used to be at 50, then it became 150 (300% improvement), then it became 300 (200% improvement), now it's 450 (i lagged to bits today, lol, but it's still a 50% improvement), and maybe i'll become laggy at 700 with the supercomputer (i'm just guessing and being negative =p). but eventually, it always will have some sort of roof.
Hieronimus Rex	Posted - 2008.07.26 02:29:00 - [6] Wow so I guess we really are ****ed. I was really hoping to be wrong in the OP. I guess then everyone who complains about lag is bored, ignorant, or a troll.
Hieronimus Rex	Posted - 2008.07.26 02:30:00 - [7] Originally by: Sovereign533 Originally by: Hieronimus Rex if it'll ever completely eliminate lag, uhm, no. the problem with lag is, it's just increasing the bar when it lags. lets say it used to be at 50, then it became 150 (300% improvement), then it became 300 (200% improvement), now it's 450 (i lagged to bits today, lol, but it's still a 50% improvement), and maybe i'll become laggy at 700 with the supercomputer (i'm just guessing and being negative =p). but eventually, it always will have some sort of roof. Yeah I know...so I guess then there's the question of how to give players an incentive (other than lag itself) to avoid hitting this roof.
J Kunjeh Gallente	Posted - 2008.07.26 02:55:00 - [8] Yes. In time, in time. And if there's one development team that I know can pull it off, CCP is it.
Veldya Caldari Guristari Freedom Fighters	Posted - 2008.07.26 03:06:00 - [9] They need to get rid of the global chat channels. Anytime some idiot says something in a channel which has 700 people in the channel the node has to send 700 packets out with the info, this is one example of server side bottleneck where it only has so many ports and only able to execute so many instructions at once. For combat related lag, again, they should have significant server side filtering. You know all that crap you filter out, it is client based filtering, the server is still sending you wads of useless crap, your client is just filtering that crap but it is still choking the server down by sending wads of useless garbage to each client. EVE should really be able to handle thousands of simultaneous in system. They just need to re-code the core of the engine which relates to information sharing and build in server side filtering and proximity need to know information filtering system. Do I really need to know what hundreds of ships many au away from me are doing? Do I really need information on what ships are doing near me that are well outside my lock range? Important information needs to be prioritised but the wads of crap sent to clients which clog up the server with even just a handful of ships is just staggering. Drones/Fighters need to be overhauled as does the missile system. It is just more crud I don't need to know about which stops the server dead. There are a lots of ways they can improve the system, some may require some in-game mechanic changes but I would rather playability be the overriding factor rather than look or feel, because something that looks or feels nice is cold comfort when you are lagged out for 10 minutes because the server is bottlenecked.
Zeba Minmatar Pator Tech School	Posted - 2008.07.26 03:37:00 - [10] Blah blah infiniband etc etc... inappropriate signature. ~WeatherMan
teh punisher Suddenly Ninjas	Posted - 2008.07.26 03:56:00 - [11] i think you hit the nail on the head. if ccp makes fleet battles stable for 300, then 400 will show up and complain. there's no limit as the game turns more and more to large fleet combat and blobbing.
Zeba Minmatar Pator Tech School	Posted - 2008.07.26 04:04:00 - [12] Originally by: teh punisher i think you hit the nail on the head. if ccp makes fleet battles stable for 300, then 400 will show up and complain. there's no limit as the game turns more and more to large fleet combat and blobbing. The upside of that is the people who run ops with well less than whatever the new limit will be will enjoy lag free battles unlike the current situation. Tbh there will never be a situation that the playerbase will not run fleets to the max or beyond the limits of the server until the underlying mechanics of fleet warfare are changed. Good luck CCP on that one. inappropriate signature. ~WeatherMan
Ackuula	Posted - 2008.07.27 02:20:00 - [13] Originally by: Hieronimus Rex Originally by: Tarminic Well, a truly dynamic system would be able to divide combat/collision calculations across nodes in the entire Tranquility cluster, but allowing multiple nodes to handle combat on a single grid is a massive technical hurdle (from what the devs have said). Eventually? Yes. Soon? No. Are we talking massive like rewriting EVE? No we are talking about a massive rewrite of how all servers work. This isn't just a problem with MMO's, all business run into problem trying trying to let multiple people have realtime access to the same databases. CCP is actually doing quite good compared to most.
Dirk Magnum Spearhead Endeavors	Posted - 2008.07.27 02:25:00 - [14] Edited by: Dirk Magnum on 27/07/2008 02:25:28 If I had to make a wager about which was more likely: 1. God exists 2. CCP can reduce the lag in Eve I would totally go with option two tbh.
F'nog Amarr Celestial Horizon Corp. Celestial Industrial Alliance	Posted - 2008.07.27 09:17:00 - [15] Originally by: Tarminic Originally by: Hieronimus Rex Originally by: Tarminic Well, a truly dynamic system would be able to divide combat/collision calculations across nodes in the entire Tranquility cluster, but allowing multiple nodes to handle combat on a single grid is a massive technical hurdle (from what the devs have said). Eventually? Yes. Soon? No. Are we talking massive like rewriting EVE? Pretty much - the server-side combat engine would have to be rewritten to be multithreaded, which is both a design hurdle in itself (multithreaded code done right ain't easy) and if I remember correctly, the limitations of using Stackless Python also play a part, though I'm not familiar enough with the language to say exactly what that limitation is. I seem to remember the devs saying that at one point they were rewriting the most computationally intense parts of the server software in C/C++, but I can't remember where I heard that. Yeah, Python and multicore are pretty much verboten. This is why CCP are trying to move to C (whatever flavor) for certian (maybe all) parts of the code. I too remember that they're moving towards that. It was probably in a Need for Speed blog, but there are so many at this point that I can't point towards the specific. But I definitely remember that an important part of the infiniband system is that C should/will become the norm. Originally by: Kazuma Saruwatari F'nog for Amarr Emperor. Nuff said Originally by: Chribba Go F'nog! You're a hero! Not a Zero! /me bows
Karanth Gallente Federal Defence Union	Posted - 2008.07.27 09:29:00 - [16] Well, if we all moved into CCP server HQ and played there, that would be fun. Might even help the lag.
Dr Slaughter Minmatar Rabies Inc.	Posted - 2008.07.27 10:03:00 - [17] Originally by: Hieronimus Rex Edited by: Hieronimus Rex on 26/07/2008 02:16:44 Other than simply pack more people into the system before it lags.... For example... Right now Jita has its own node and can pack 700 or so people in with a fair bit of lag. This means that when there are 700 people the 701st person is contemplating entering Jita but chooses not to because of lag concerns (often I am such a person on Sunday afternoons). If 700 people could be in Jita on Sunday with no lag, there would probably be more people in Jita, say 1500 (myself included), until it lagged again. Same idea with fleet battles... the 201st person considering going on a fleet op might choose not to because of lag worries. If CCP attempts to combat lag blobs will only get bigger. Of course plenty of "ohlol it's lagging in Motsu, Jita, and fleet battles" threads are simply trolls or idoits, but presumably there is a nonzero number of threads posted by people who really actually believe CCP can do something about lag. I'm looking for their input. To really understand this topic start out by reading the majority of: this thread The short answer is lag will be with us for a very long time when it comes to combat as people while people are allowed to bring as many players as they can to a grid. BUT... as CCP implement infiniband and re-write the sol services we should see incremental improvements. The issues around Python/C++ come back to Pythons Global Interpreter Lock (GIL) which acts as a bottle neck as it stops you being able to take advantage of multi-core design to it's fullest (simplification, but if you want to read more google Python GIL). Re-writing in C++ removes the GIL issue for that code but not the rest and you add over heads switching back and forth between. So long and short appears to be: Gradual incremental improvements as they roll-out infiniband. Some areas will become much more 'snappy' (grid/system loads) as they can move a ton of data between physical hosts much quicker. As fights get bigger.. lag will always increase. and it would be very interesting if CCP could get hold of one of the 6Ghz Power-6 systems from IBM and try running stackless on it.. the reason they're in the current situation is that they assumed the Ghz of CPUs would keep increasing but Intel and AMD decided to go multi-core instead (which CCP have to re-write for to support properly).. IBM however have carried on cranking up the Ghz so, in theory, look like a great solution for running CPU bottlenecking Python code. meh. whatever.
kyoukoku	Posted - 2008.07.27 11:28:00 - [18] Originally by: Dirk Magnum Edited by: Dirk Magnum on 27/07/2008 02:25:28 If I had to make a wager about which was more likely: 1. God exists 2. CCP can reduce the lag in Eve I would totally go with option two tbh. ^^This!
Wild Rho Amarr Silent Core	Posted - 2008.07.27 11:37:00 - [19] Originally by: Tarminic Pretty much - the server-side combat engine would have to be rewritten to be multithreaded, which is both a design hurdle in itself (multithreaded code done right ain't easy) and if I remember correctly, the limitations of using Stackless Python also play a part, though I'm not familiar enough with the language to say exactly what that limitation is. I seem to remember the devs saying that at one point they were rewriting the most computationally intense parts of the server software in C/C++, but I can't remember where I heard that. The problem with spreading processes out over multiple threads is it costs resources to manage effectively (making sure many processes are allocated to one thread with only a few allocated to another) and you also have to make sure combat calculations are all carried out in the correct order or you could find yourself being shot down by a ship that was already killed or had warped out 5minutes ago not to mention the potential for errors. The overhead, added complexity and potential for things to go horribly wrong just aren't worth it. Stackless python also doesn't do multi-threading if I remember right.
bloomich Caldari In Siders	Posted - 2008.07.27 12:01:00 - [20] Non-technical question: "How far are we towardes a solution to lag"? Are we still at brainstorming stage? At designing a solution stage? Implementing a solution? Testing one? Year away? 5 years away? etc etc. These are the questions we need to ask. Not armchair technical ones.
Stab Wounds Caldari State Protectorate	Posted - 2008.07.27 12:04:00 - [21] use the lag against your enemies
Cor Aidan KNIGHT'S OF THE ROUND ROOM ReZZerecteD AlckemisTs	Posted - 2008.07.27 18:16:00 - [22] You can never win against lag. *Every time you double the number of people you want to interact in an area, you have to quadruple* the computing power and communications bandwidth.** There is no theoretical way to eliminate this problem, let alone an actual technical solution. Multithreaded capability has nothing to do with it. In fact, you will eventually reach the point where no matter how good the CCP server node is, you just can't get enough information over your personal network connection to handle all the information.
Rawr Cristina Caldari Omerta Syndicate	Posted - 2008.07.27 18:21:00 - [23] CCP just need to find a way to spread people out imo. If the server can't handle 600+ people in 1 system then give people a reason to spread out like, congestion charges ...
Artemis Rose Eleckrostatik	Posted - 2008.07.27 18:22:00 - [24] Go to low sec, stay out of the really crazy hotspots. OMG all your lag troubles are gone. __________________________________________________ Currently Playing: Trolls from Outer Space Current Equipment: VISAcard chain mail, +2 Amulet of Epic Whine. WTB Purple Nerf Bat.
bloomich Caldari In Siders	Posted - 2008.07.27 18:31:00 - [25] Edited by: bloomich on 27/07/2008 18:34:18 Originally by: Cor Aidan You can never win against lag. *Every time you double the number of people you want to interact in an area, you have to quadruple* the computing power and communications bandwidth. There is no theoretical way to eliminate this problem, let alone an actual technical solution. Multithreaded capability has nothing to do with it. In fact, you will eventually reach the point where no matter how good the CCP server node is, you just can't get enough information over your personal network connection to handle all the information. Incorrect. Back in the days of 600 baud, people like you stifled the development of major advances. Remember the days of people like you saying it was impossible to have development of more cyllanders on engine blocks. Or impossible to transmit more data through a copper wire. Or impossible to have a ship of metal to float. Or impossible to fly. The solution is not impossible forever. It may not be practical at this time. Even your theory of doubling people = quadrupling computing power and bandwith is incorrect as well. So forget all this rubbish about Servers and nodes etc being impossible. Let us talk about reality. The Real question we should be asking is how far along the line is CCP towards implementing a solution to reduce or eliminate lag?** Are we still at the drawing board? Are we aready testing something? Are we looking at a estimated 5 years? 2 years? 6 Months? 10 Years?
Cor Aidan KNIGHT'S OF THE ROUND ROOM ReZZerecteD AlckemisTs	Posted - 2008.07.27 19:35:00 - [26] Originally by: bloomich Edited by: bloomich on 27/07/2008 18:34:18 Originally by: Cor Aidan You can never win against lag. *Every time you double the number of people you want to interact in an area, you have to quadruple* the computing power and communications bandwidth. There is no theoretical way to eliminate this problem, let alone an actual technical solution. ... ... The solution is not impossible forever. It may not be practical at this time. Even your theory of doubling people = quadrupling computing power and bandwith is incorrect as well. ... Your other points are based on this incorrect assumption. Communications requirements going as the square of participants is not a "theory", it is a fact (I'm not going to show the math again, as it's been posted numerous times in these forums). That said, in the context of EVE, the following is the only solution: Originally by:** Rawr Cristina CCP just need to find a way to spread people out imo. If the server can't handle 600+ people in 1 system then give people a reason to spread out like, congestion charges Well, I suppose, that is, until we have developed new computing and telecommunications infrastructures that scale much faster than the current methods, but I suspect that will be up to entities other than CCP to develop and, assuming CCP is still around, then CCP may adopt those new technologies.
Gnulpie Minmatar Miner Tech	Posted - 2008.07.27 21:57:00 - [27] Originally by: Cor Aidan Originally by: bloomich Originally by: Cor Aidan You can never win against lag. *Every time you double the number of people you want to interact in an area, you have to quadruple* the computing power and communications bandwidth. There is no theoretical way to eliminate this problem, let alone an actual technical solution. ... ... The solution is not impossible forever. It may not be practical at this time. Even your theory of doubling people = quadrupling computing power and bandwith is incorrect as well. ... Your other points are based on this incorrect assumption. Communications requirements going as the square of participants is not a "theory", it is a fact** (I'm not going to show the math again, as it's been posted numerous times in these forums). Oh yes, it is a theory and that theory is correct only under exactly defined assumptions. And one assumption is that everyone on the grid needs to communicate with everyone else. THEN you get indeed quadratic runtime or worse. However, nowhere it is stated that you need to have one huge grid where everyone needs to communicates with everyone else. The really big problem is that people are so fixed on the mantra which you are spreading also that they don't think about alternatives where not everyone has to communicate with everyone else while still the illusion of freedom is maintained! THAT would be worth some major research because here you can gain massively. Imagine for example that there are groups and people would communicate to groups and only actively communicate with people inside exactly one group at a time while from the other groups only the group status is transmitted. Imagine a group size of 10 people. If you have 150 people then everyone needs communication with those 15 groups (15 groups a 10 people = 150 people) plus one group of special interest. So, 10 people from the special interest group + 14 remaining groups = 24 communication. 24 instead of 150 in the basic approach. For one person. For all the people together it would be easily calculated 24150 vs 150150, that is 3600 vs 22500. If you have 300 people you would need communication with 10 people + 29 groups = 39 instead of 300. Do the math for the whole group again and you have 11700 vs 90000. BIG savings. That is of course only a rough idea and there are ptifalls left and right. But THAT would be some worthwile approach for research. And it shows that the naive assumptions that the communication costs are the square of the people on the grid is just too simple. Sure, the square of the participants. But as said, what if you have groups of people as participants instead of single people? Theories are only valid under certain assumptions. Tweak the assumptions and you will get new possibilites.
GallenteCitizen20080615 Gallente Federation War News	Posted - 2008.07.27 22:01:00 - [28] yep i do, as it the people causing the lag....once the server belongs to me i will have a dedicated area for myself to do what i like in it...ohhh wait i already have the jovian homeland ====D we will have to see once inifiniband comes out i believe tghere will be a jerastic improvement in lag reduction but once people know it stop lagging they will go to jita in more numbers but fleet fight should be more dooable i believe other than that lag doesnt effect me much as A. i dont mission run B. i dont go to jita C. i only get lagged out with 100 vs 100 battles of more which i do roaming gangs Originally by: CCP Wrangler We are pleased to aim! Or was that the other way around?
Chainsaw Plankton IDLE GUNS IDLE EMPIRE	Posted - 2008.07.27 22:31:00 - [29] Originally by: Joakim Wasyl I don't follow the, double the people quadrouple the procesing power bit. I only have 8 guns, so I can only affect 8 other people on the grid surely no matter how many you add ? It's not like we are all interacting (where you would get some kind of exponential increase). no, but if you get blown up, that information has to be sent to everyone else. likewise if you are changing position that also has to be relayed to everyone else. if anyone has you targeted your hitpoint info has to be sent out. and likewise you want to be able to see where everyone else is.
Wild Rho Amarr Silent Core	Posted - 2008.07.27 22:32:00 - [30] Originally by: Joakim Wasyl I don't follow the, double the people quadrouple the procesing power bit. I only have 8 guns, so I can only affect 8 other people on the grid surely no matter how many you add ? It's not like we are all interacting (where you would get some kind of exponential increase). You're forgetting you need constant updates on the targets position and movement, whether they're firing or not (so your client can spawn weapon effects or missile objects) and so on. For each new pilot that new pilot needs updates on what everyone else is doing and everyone needs updates on that pilot hence the growth. For example... 2 pilots on grid A is updated about B B is updated about A 3 pilots on grid A is updated about B A is updated about C B is updated about A B is updated about C C is updated about A C is updated about B ...and so on. As you can see the amount of updating required increases considerably with each new pilot added.

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.