Comments

Over the next few days, I want to wrap up a whole series of posts on a variety of topics, most of them centered on those last few features of Rubicon that I haven’t yet talked about, plus a few things that came up at our Town Hall meeting on Saturday.

That starts with a question that I was both asked directly recently (twice!) and which also came up in the Town Hall: “What do I think about Time Dilation?” And my position on TiDi is fairly simple: I think it’s a brilliantly designed, brilliantly implemented band-aid that I wish was not necessary to the survival of EVE Online.

Now granted, TiDi is immensely superior to what came before: black screens, unpredictable behavior, massive lag, et cetera. TiDi is better than all of that. No question. But is it introducing its own problems?

If CSM8 Chair Trebor Daehdoow had a law, it would be “Fleets expand to fill the lag available.” Back when TiDi was first introduced to the game, I predicted that it would simultaneously be the best single change ever for the game and the worst piece of garbage ever to be inflicted on the game. These were not incompatible beliefs. Enormous fights are good for EVE Online: they great for the game’s publicity and press, great for marketing and promotion, and contribute to the epic feel of the game. But I also predicted that it would simultaneously increase alliance sizes such that they could put thousands of ships on the field if they wished to attack or defend a position.

Both of those predictions have come to pass: enormous fights have been successfully used for marketing EVE Online and alliances sizes and fight sizes have doubled since TiDi was introduced with no corresponding increase in PCU count. Trebor’s law has been coming to pass: since alliances could put thousands of ships on the field, they have been.

And with these changes, the underlying problem that caused TiDi to be added to the game in the first place is again starting to rear its head. We’ve all now seen TiDi inflicted on us during routine in-game situations. Alliances are now large enough that they can reinforce 15 targets at once across multiple regions and then pick and choose which of those 15 they’re actually going to attack when the timers hit. They’re even big enough to show up to attack more than one of those targets simultaneously if they choose to. That makes asking CCP to reinforce the right systems tricky or impossible. Does the attacker or defender ask for all 15 systems to be reinforced? Even when the node is reinforced, some TiDi happens anyway.

TiDi is a brilliant band-aid and for now, it’s still holding the wound closed. But yeah, we can now see it’s not going to hold forever.

At the Summer Summit, I and other members of the CSM warned CCP Seagull directly that her vision for EVE, while fascinating, promises to create the biggest timer in the history of EVE Online. If a few thousand people wanted to be in Asakai for the fight there, can you imagine how many people are going to want to be in the system to see the first player-constructed stargate brought on-line… or not, depending on how the fight over this goes? I certainly want to be there, and I suspect there will be thousands and thousands of others that will be too, enough to make any prior gathering of players in a single system look like a faction warfare skirmish.

EVE is soon going to need more serious band-aids or — preferably — it’s going to need an architectural redesign at some fundamental level to handle bigger and bigger fights. This isn’t the job of one team; this is soon going to be a challenge facing the entire company.

In the meantime, TiDi is causing problems that couldn’t have been predicted back in 2011. At that time, I wrote:

In order to survive in 0.0, you have to learn, develop, and practice a whole skill-set around managing deficiencies in the so-called “game” you are playing.

And that’s swiftly becoming true around TiDi as well! One of the key components that made the recent live event so deadly for high-sec players, after all, was the fact that the null-sec alliances living around Doril had something like two hours to prepare their kill box… thanks to TiDi. Had TiDi not been a factor, the kill box wouldn’t have been nearly as efficient and the event might well have gone differently.

So now TiDi’s not only impacting the game, it’s creating an all-new meta-game, an emergent game-play mechanic that didn’t exist and couldn’t exist without TiDi! I’m pretty sure that wasn’t the intent. But it’s interesting, don’t you think? Discuss.

– Ripard Teg

53 Comments

  1. Provi Miner

    derp have posted the exact same thing dozens of times. We need a hardware update of the most serious kind. even with tidi nodes crashing have affected the outcome of at least three huge block battles recently and this is unacceptable.

    November 18, 2013 at 4:39 pm Reply
    1. Duh

      Maybe if uninvolved turdlords would stay out of it this wouldnt be an issue. Lookin at you Alphastarpilot and RVB, Rooks & Kings, Bombers Bar, Outbreak and other scrubs who leach off the big boys content.

      November 18, 2013 at 4:55 pm Reply
      1. bob

        If you want fair play you are playing the wrong game.

        November 18, 2013 at 5:08 pm Reply
        1. Duh

          Didnt say anything about fair. This is an article about tidi and how every scrublord from highsec gets in an Ibis and comes to whore on some mails when something big pops off. So put that in your small-gang and smoke it mister.

          November 18, 2013 at 5:22 pm Reply
          1. Kamar Raimo

            Awww
            Did someone crash your 1000 VS 1000 blob party by bringing another 50 guys or did R&K pipebomb your fleet?

            Maybe you’d like a separate server only for blob fleets? ; p

            November 18, 2013 at 6:52 pm
          2. whaahaha

            awww did someone come off mittens cock long enough to post on a shit site?

            November 18, 2013 at 6:59 pm
      2. troll

        You sure are a big boy who can push F1, have a cookie. Now get off the l33t-nullbear-sperg-horse and get it into your empty windy skull – EvE is a sandbox open to everyone, deal with it or GTFO.

        November 18, 2013 at 5:16 pm Reply
        1. slo

          U mad bro.

          November 19, 2013 at 1:20 am Reply
      3. orif

        if you N3 shitcunts would just leave already to FW we wouldnt need to fight in tidi.

        November 18, 2013 at 6:20 pm Reply
        1. anonymous

          DUH is Black Legion moron.

          November 19, 2013 at 2:40 am Reply
    2. mooo

      the columnist is right… even if you update the hardware, you will still have a problem because instead of 4k fighting in a system, you’ll have 8k. Not only the major coalitions have tens of thousands of players, but sometimes they even ally with each other. I don’t really see a solution for this problem, unless there was such a major update in tech that it would be possible to have almost all of the players in the same battle at the same time. Doesn’t look likely, even if CCP had Google pockets…

      November 18, 2013 at 5:22 pm Reply
      1. Provi Miner

        The solution is already there just needs to be updated (and thats where hardware comes in) most crash’s occure because of ccps inability to isolate systems quickly enough (as in once a system hits 80% tidi it should auto kick that system to an unused server) Thats a hardware issue more than a software issue. BTW in case no one noticed durring the live event something odd occured, despite 10% tidi all around the core systems never got worse then 50% even with enough pilots and ships to cause 10% on a single server :) said it before the live event looks like a hardware trial run and it was partially successful.

        November 18, 2013 at 6:58 pm Reply
        1. Surloc

          That won’t work. If you kick a system to another hardware server you have to boot everyone offline in that system. The migration doesn’t happen live. If you do that a there is sizable chance the losing side will simply not log back in. What you can do is start kicking the other systems on that server, but that means you are downing people there.

          November 18, 2013 at 7:50 pm Reply
        2. Playos

          Ya, everything you just suggested… is software, and major software, chages. Nothing you stated talked about what the hardware does. Increasing the clock speed of the CPU, the amount/speed of the RAM, or the any other point of hardware is going to fix the problem with what CCP has coded today…

          With that said, unraveling the code base and making it easier to maintain and more modular is the first step… so we’ve had that going on for about 2 years and we’ve got 3 major areas that haven’t been touched on (market, overview, and roles/rights) and another that will need another pass at the back end to catch up with the changes since it was the first (inventory). Problem for CCP is that none of these things are all that front end visible… they aren’t bullet points for expansions and the end result will just be the start of a major undertaking (migrating some of those now modular systems to multi-threaded designs).

          November 22, 2013 at 4:51 am Reply
  2. Csm8 is shit

    Where’s the fucking minutes

    November 18, 2013 at 4:48 pm Reply
  3. RBW

    I have not read any of your articles, but after reading this one I feel it worth it to find them and read up. I like the style of your writing. Thank you.

    November 18, 2013 at 4:49 pm Reply
  4. Some guy

    Back then, TiDi was introduced to make the big fights work better. The problem now is that the fights have become bigger and more regular. Even if there are some people who will always get a boner to be in such battles, I think they frustrate more players then anything. On the long run, these TiDi/lag fests will have a negative effect on null sec player population. At some point, CCP has to ask itself: Who wants to be on 4 hour long fight on a regular basis? So what should CCP do to reduce this? Limit fight size on timers? Nerf coalitions? Everything should be on the table. Including the things that will make some player rage. Believe me, I dont like these solutions. But at some point, something has to give.

    November 18, 2013 at 5:33 pm Reply
    1. Yoji

      FIghts have always become bigger. Some of use remember when fleets of 20 people would generate lag. Then it was 100. 200. 500. Then first time moor then 1000 players were in system, people said, this is it. Now we have 3000+

      As long as 0.0 is not fixed. People will swamp the system and fuck everyones game.

      November 18, 2013 at 6:55 pm Reply
    2. Provi Miner

      I like the idea of nerfing coalitions but you would have to change the map, there are only so many entrances and exits that can be used right now. Every null region should have 2 to 3 low sec systems between it and high sec and the null regions should be made small and more of them and have them fully buffered by low sec from each other.

      November 18, 2013 at 7:03 pm Reply
  5. anon

    Player-constructed stargates? Is this a leak in violation of the NDA or have I been missing some devblogs?

    November 18, 2013 at 6:02 pm Reply
    1. Duh

      It was in the keynote at fanfest…

      November 18, 2013 at 6:04 pm Reply
      1. WatermelonChickenJesus

        Just like walking in stations was a keynote.
        It will never happen.

        November 19, 2013 at 10:35 am Reply
        1. Playos

          Did happen, captain’s quarters… and now it’s called DUST

          November 22, 2013 at 4:41 am Reply
  6. Boris

    A bit has changed since 2008, infiniband increased from 2.5Gbps to 56Gbps for example. Perhaps time to open up the wallets and spring for another upgrade?

    November 18, 2013 at 6:43 pm Reply
    1. no

      Bandwidth is not the issue.

      November 18, 2013 at 7:38 pm Reply
    2. Surloc

      The problem is not interconnect between hardware nodes as much as it is the speed of a given hardware node. Eve is not designed to run a system on more than one piece of hardware. It’s a matter of raw cpu speed. The problem is cpus aren’t getting faster on the high end. Sure they are giving you more cores, and more efficient cpus, but in the end even the best multithreaded apps tend come down to a single thread which is the bottleneck. Once that happens you simply need a faster cpu not more cores, or nodes. Sure you can rework things to remove the bottleneck, but you’ll simply hit another bottleneck some where else.

      PS- IB was at 20-40Gbps in 2008. Which is not to say CCP couldn’t benefit from Mellanox nics and switches running at 56Gbps with ethernet or IB on the back end…

      November 18, 2013 at 7:45 pm Reply
  7. Shadowblade1436

    I have always argued that TiDi could be used in a different way for sov war.
    Drop SBU’s in multiple systems, reinforce all of them, wait till the time, drop in a HUGE fleet to one, the defenders drop in there HUGE fleet to defend and TiDi happens. Now while everyone is stuck in TiDi hell for the next 4 hours, you uses a second smaller fleet to go finish off the other lesser important systems and online your TCU, even if you loses the main system, you still took a ton of space.
    But would that be an exploit of game mechanics? Its not the way TiDi is meant to be used according to CCP. Would CCP be forced to return all the systems back to there old owner? I think TiDi is a sleeping monster that has yet to wake up, and when it does CCP will have a shit storm to deal with.

    One thing CCP should be looking into is programming some kind of automatic TiDi detection system, that will isolate the TiDi, then begin to reinforce the Nod on the fly, Removing more and more other systems off that nod, as TiDi goes up. And start looking into better servers to handle 2000+ man fights, with out having to go to 10% TiDi seeing that kind of fight is now common in Null sec and isn’t going away anytime soon. Even is it means increasing the cost of subscriptions to from $14.95 to $16.95 to pay for new servers. I think people will pay a few more dollars if it meant TiDi didn’t even start till you hit 2000+ in a system.

    November 18, 2013 at 6:49 pm Reply
    1. Rekkr Nordgard

      That strategy of pinning your enemy’s fleet down in a TiDi system with a diversionary force and then hit other places with another fleet has occurred to me as well and I’ve always been curious as to why no one has tried it yet. Even if it only works once, it could still get momentum rolling your way and make the enemy less eager to commit it’s full force in the future possibly giving you an edge.

      November 18, 2013 at 8:30 pm Reply
    2. Ashesofempires

      CCP already has some of the best servers they can buy. At this point, there isn’t much left for CCP to do in terms of hardware. It’s entirely limited by the IPC (instructions per clock) of each thread of the processor. Each node is a single thread, and would benefit only a tiny amount from the 5 or 6% improvement in clock speed from buying all new hardware. In fact, because of “market segmentation” policies, the lower end processor parts from AMD and Intel, while having fewer cores, are usually clocked slightly higher (due to thermal constraints on the more expensive parts) which means that the cheaper processors are usually a better choice for CCP’s purposes.

      The only real path forward is one CCP is unwilling to dive into: making EVE’s server backend multi-threaded. This involves ditching Python and rewriting almost all of the game’s backend code in some other multi-threading friendly development language.

      November 18, 2013 at 9:34 pm Reply
      1. red

        lol not true. Nvidia has designed servers that will put ccps entire room of servers into one. gf

        November 18, 2013 at 9:46 pm Reply
        1. Grunt

          If you work in IT, please quit your job and go shovel or something. That is one answer no real IT person would give.

          November 19, 2013 at 9:16 am Reply
        2. Ashesofempires

          No, they have not. They, just like the rest of the server hardware engineers, have developed hardware designed for threaded processes. Only more so; the servers designed by nVidia are designed for either massively parallel threaded operations or for tasks that don’t individually require a lot of processing power. Their servers play to the strengths of the CUDA architecture. Lots of threads, not very fast. Great for some things, but not for a server process like EVE’s. EVE’s backend is nearly as serial as a program can get, thanks to the python language’s restrictive global interpreter lock, which prevents more than one thread from manipulating the same variable.

          November 19, 2013 at 5:07 pm Reply
  8. Guest

    Can’t they come up with a more innovative way to reinforce nodes than turning off 3 of 4 cores and using them as a heat sink?

    I know processor development no longer supports high speed single threaded architecture, but there has to be a better solution.

    Are there no experimental single core, high speed, next-generation cooled processors available? Could we perhaps retrofit a more specialized processor for this application?

    November 18, 2013 at 8:01 pm Reply
    1. Ashesofempires

      The reason why everyone transitioned to multi-core architectures was that everyone came to the same conclusion: past a certain operating frequency, there’s too much leakage of electrical current, which causes a phenomenon where molecules are stripped from the transistors and deposited elsewhere. This is in addition to heat generation, which was getting to the point of absurdity at the end of the single processor era. (Pentium 4, IBM’s Power architecture)

      If CCP’s past development blogs on the subject are at all true, then there exist a few ways to solve the problem:

      The first is more bandaid-like, but probably easier. Move more non-combat related processes away from the nodes and to their own nodes. The end result would be that only things like positional data, damage calculation, and such are being handled by the node. All other API calls, like mail, station services, inventory, etc are being handled by another node. This was already done to some extent. CCP found that they couldn’t easily separate inventory from the local node, because it caused problems with things like deploying drones and reloading guns.

      Second, CCP could move away from Python and its shitty Global Interpreter Lock into a develelopment language more multi-threading friendly and invest some serious cash and man-hours into revamping the entire backend of the game to move it into the modern era of software programming. I understand why they don’t (it’s a monumental undertaking), but they are laboring to fix a problem that was caused by building a massively multiplayer game in the “golden age” of Intel insisting that 10ghz was just around the corner, and single threaded designs were never going away.

      Third, CCP could go beyond just multi-threading the node, and revamp the grid system itself to do two things: First, the grid itself would get larger: 10,000 kilometers by 10,000 kilometers. This is large enough to fix the problems caused by a 250x250km grid that we see now, but not include multiple celestials (stations, etc) in one grid. Second, they make each grid its own node-within-a-node. This gives some granularity to the system. Now each individual grid can support a large number of players, while the system node itself only has to worry itself with pilots that are in transit between grids, docked, or whatever. CCP could do a few things with these new grids. They could place them at tactical locations (POSes, Stations, etc), or have them spin up more dynamically. The first option is more feasible, as the vast majority of mass fights happen in the space around static structures, but leaves open the possibility of shitty lag in fights that happen spontaneously away from a static structure (like Asakai, or any fleet fight that occurs on a drag bubble). Dynamic grids are harder to do; it involves determining at what point players are moved onto a dedicated node, and how to do it without causing disconnects or other problems for the players.

      I’m a big fan of just getting rid of Python and allowing a node to utilize more than just the one logical thread of a multi-core processor. I would imagine that once CCP got the synchronicity problems solved, the power of an 8-core, 16-thread server processor would help tremendously. Or even Intel’s upcoming 60-core co-processor…

      November 18, 2013 at 9:25 pm Reply
  9. Boliano

    I think the solution is easier. Make it harder for people to move 20000 people from one sid of eve to the other. Lets face it someonelike cfc. N3 can easily move tens of thoussands of players with titan bridge and jumpbridges. Get rid of them. If ur northern eve coalition wants to attack the south. You should have to work ur balks off getting there. In rl if say the U S wants to attack russia the logistics to do that is insane. And extremely expensive. Its too easy to move region to region with a zillion people. Then these major entities will expect there holding alliances to defend more themselves without including half of null in every singke fight.

    November 18, 2013 at 8:25 pm Reply
    1. M1k3y_Koontz

      I support this message and/or product.

      Titan bridges are too easy, and removing them would be the least of the evils.
      By forcing blobs to move by gate, 2000 people moving into a system would have to move in 8 groups, slowing down larger fleets and allowing smaller groups the chance to complete their objective before the larger group arrives.

      November 18, 2013 at 10:09 pm Reply
      1. kill yourself

        yea because hundreds of people jumping gates everywhere wouldnt cause any problems…fucking idiots…

        November 18, 2013 at 11:22 pm Reply
        1. Dumbcunt

          Well maybe nobody would bother anymore to jump 100s of people through gates in that case and would instead have those people engage in more satisfying PvP endeavours?

          November 19, 2013 at 12:19 am Reply
          1. oh eve players

            a solution to a problem that is worse than the problem itself …..your name…so fitting..

            November 19, 2013 at 6:44 am
          2. WatermelonChickenJesus

            plain stupid.

            November 19, 2013 at 10:34 am
        2. Mikey

          The point is that there WONT be hundreds of people jumping gates, tidi can handle 250 people without too much trouble but 2000 people wont be able to handle it

          November 19, 2013 at 11:11 am Reply
        3. Maire Ford

          Calling people names makes your point of view automatically valid and gives you credibility. I like that! Please continue. It entertains me.

          November 19, 2013 at 2:37 pm Reply
          1. WatermelonChickenJesus

            he is not calling peoples names he is calling them by what they are…
            fucking idiots.

            November 19, 2013 at 2:52 pm
    2. Dumbcunt

      Nerfing force projection would solve so many balancing issues in a single sweep. Sadly those tards at CCP don´t recognize this because a 4000 man lagfest gives them media attention.

      November 18, 2013 at 10:31 pm Reply
      1. Uknowwho

        CCP marketing guy – I had sex with identical triplets last night.

        Reality of Tidi – True…but they were each 350 pounds and smelled like fish and ciggarettes.

        CCP is a great example of losing control of your marketing group. EVE turns into pong on Quaaludes when any substantial numbers engage but CCP continues to market EVE the ability to bring thousands together in combat. It would be interesting to see them try to portray EVE correctly showing 10 % TIDI.

        Most people that play EVE are quite bright. That said, I’m always amazed at their ability to ignore how terrible EVE’s combat system is. Eve is a stunningly beautiful, brilliant game – until you want to kill someone then it’s as fun as signing up for Obamcare.

        November 20, 2013 at 2:38 pm Reply
    3. WatermelonChickenJesus

      which will in the end stagnate null sec even more.

      Very short viewed solution only recommend by idiots.

      Cold as Ice

      10 hours ago
      Make titans have to jump w/ the fleet they bridge fixes a few things

      ^^^^^^
      seems way smarter.

      November 19, 2013 at 10:33 am Reply
  10. Cold as ice

    Make titans have to jump w/ the fleet they bridge fixes a few things

    November 18, 2013 at 11:58 pm Reply
    1. smallcaledude

      good idea for the large scale , but i fear small groups (with 50 people on field an less) would suffer from this … wich again doesnt mean its a bad idea necessarly

      November 19, 2013 at 12:06 pm Reply
    2. Dreamer.

      Yes it could be one way to solve parts of the problem. Another could be if the the cyno gets popped while your in transit, then you end up in random system in range of the titan …

      now that would mess up your fleet alot. scattering it across several systems.

      But the real issue is not the Titan bridges my real issue is the Huge renter empires spread out. this is the real plague of Nullsec in my humble opinion

      If they make so the Cost of the Alliance to band together is based on number of corps in the alliance.

      and the cost of sov for the holding corp is X ^ (system_held) then we get a natural spread of the systems. among the corps and still restricting the alliance from creating 20x Holding corps.

      November 19, 2013 at 3:54 pm Reply
      1. Cold as ice

        Titan should be-able to bridge x number of ships, also x number of light-years.
        Force projection is the problem of the game.

        An alliance should not be able to go to one side of eve universe to the other in 15 minutes. Moving to location X. Should be a commitment not a 3 hour tour.

        Renter alliances. The cost of sov goes up X per system owned per system
        so the more systems you own the more the sov costs you.
        This will make these giant renter alliances less attractive. and lots of small renter alliances harder to maintain.

        the biggest problem of the game, force projection and the amount of isk that these renters/ collations can gather.

        November 19, 2013 at 4:48 pm Reply
      2. Vandar

        Good one – Blocks would then just divide their pleb renters into smaller groups to reduce costs instead of having super renter alliances… or more likely they would just charge pleb renters more for their space.

        Making Titans jump with the fleet – this works both ways.

        – You either have less bridges and less content
        – Bridge 2 systems over and slowboat, titan cloaks up GG
        – More capital escalations, lures people to system like flies to Shit – thus normally ending in a retarded TiDi + stalemate, the node crashing then everyone surviving.

        November 20, 2013 at 12:46 am Reply
  11. Tiago D'Agostini

    Just give lots of smaller targets on infrastructure that small fleets can disable (like 20 people) therefore making more interestign to execute 40 20 man gank raids into infrastructure resources than a single 800 man fleet attack on a single target.

    Players will always go for the most effective way, make smaller engagements become the msot effective way of winning, and players will do it.

    November 20, 2013 at 3:31 pm Reply

Leave a Reply to no Click here to cancel reply.