Originally by: Phillis Stein

---

Im not experienced in PCs but I always wondered why they dont revert all graphics to polygraphics once a system hits a certain number. Have the ships shown with an just an outline as they did in Elite.

And before ppl say that that would ruin eve for them, no nice graphics etc, in a fleet fight we all zoom out anyway so a loss of graphics wont make any difference to us.

As I said, Im no PC expert so amybe this wouldnt make any difference at all

---

Originally by: McDonALTs

---

Originally by: Phillis Stein

---

Im not experienced in PCs but I always wondered why they dont revert all graphics to polygraphics once a system hits a certain number. Have the ships shown with an just an outline as they did in Elite.

And before ppl say that that would ruin eve for them, no nice graphics etc, in a fleet fight we all zoom out anyway so a loss of graphics wont make any difference to us.

As I said, Im no PC expert so amybe this wouldnt make any difference at all

---

They already do. Main issue with fleet lag is

1. Client side Poor UI taking up massive resources. Brackets help reduce this
2. Slow servers

---

Originally by: Ruze

---

Or are you saying that the servers CCP has, the ones that are the 'latest tech not even on the market yet' servers are slow?

---

Originally by: Benilopax

---

The Server.
So Hammerhead told me that the supercomputer server is beng worked on but as you can guess its a huge task to undertake and also its a supercomputer for a computer game so it needs new exciting code and other alterations and the best he could give me is Soon TM.

---

Originally by: TheEndofTheWorld

---

1. CCP fixs lag for 600-1000 man in local.
2. Players start complaining about lag with 1000-2000 in local.

---

Originally by: Phillis Stein

---

Im not experienced in PCs but I always wondered why they dont revert all graphics to polygraphics once a system hits a certain number. Have the ships shown with an just an outline as they did in Elite.

---

Originally by: TheEndofTheWorld

---

1. CCP fixs lag for 600-1000 man in local.
2. Players start complaining about lag with 1000-2000 in local.

---

Originally by: Elektrea

---

Originally by: TheEndofTheWorld

---

1. CCP fixs lag for 600-1000 man in local.
2. Players start complaining about lag with 1000-2000 in local.

---

i'm a tard

---

Originally by: Benilopax

---

So first, and I know this isn't about Lag but its worth mentioning. CCP has split into 3 departments: Outside Station (Spaceships and spacey things), Inside Station (The Market, Ambulation, refining etc etc), Bug fixing and Server (Obvious).

---

Originally by: Ruze

---

Another interesting concept, which I hope to God happens but will probably **** off many a camper or campee ... will being docked in station remove a player from the 'local' channel and display? Man, that would be wonderful, and highly annoying to station campers and those who hide within stations.

---

Originally by: CCP Hammer

---

I'll see if I can get someone else to comment but no promises.

---

Originally by: Roguehalo

---

Can you nano Titans?

---

Originally by: CCP Hammer

---

Originally by: Benilopax

---

So first, and I know this isn't about Lag but its worth mentioning. CCP has split into 3 departments: Outside Station (Spaceships and spacey things), Inside Station (The Market, Ambulation, refining etc etc), Bug fixing and Server (Obvious).

---

Let me clarify that one a tiny bit. It's not CCP that was split into 3 departments but rather the EVE Software Group that we've grouped into 3 themes. Within those themes the programmers then specialize further.

And to answer another comment that they'll "believe it when they see the hardware" I'd have to say that's the most reasonable reaction to have when faced with these sorts of claims. I will say however that there are a lot of things we need to do on the software side before we reap the benefits of Infiniband and HPC. I'd love it if we could just plug the stuff in and see massive performance gains but that's not how it works.

One final note is that I'm just a game designer and while I have a passing knowledge of the technical stuff mostly through osmosis or lies on Wikipedia I'm no expert by any means and most of it seems like black magic to me. I'm sure many of you out there have more technical questions about this stuff but I'm the wrong guy to ask. I'll see if I can get someone else to comment but no promises.

---

Originally by: Elektrea

---

Originally by: TheEndofTheWorld

---

1. CCP fixs lag for 600-1000 man in local.
2. Players start complaining about lag with 1000-2000 in local.

---

your a tard

---

Originally by: Thrawnfl

---

you forgot to mention everyone's most hated bug/feature, The Desync which ccp says doesn't exist, the one thing that has caused the most deaths of super caps.

---

Originally by: CCP Hammer

---

And to answer another comment that they'll "believe it when they see the hardware" I'd have to say that's the most reasonable reaction to have when faced with these sorts of claims. I will say however that there are a lot of things we need to do on the software side before we reap the benefits of Infiniband and HPC. I'd love it if we could just plug the stuff in and see massive performance gains but that's not how it works.

---

Originally by: Doc Fury

---

So... approximately how many lines of code have you guys "done on the software side" so far? In other words, how much EVE server code has already been rewritten for Infiniband?

I'm guessing this stuff is all still up on whiteboards and exists only in the realm of "possibilities" so far.

---

Originally by: CCP Hammer

---

Originally by: Benilopax

---

So first, and I know this isn't about Lag but its worth mentioning. CCP has split into 3 departments: Outside Station (Spaceships and spacey things), Inside Station (The Market, Ambulation, refining etc etc), Bug fixing and Server (Obvious).

---

Let me clarify that one a tiny bit. It's not CCP that was split into 3 departments but rather the EVE Software Group that we've grouped into 3 themes. Within those themes the programmers then specialize further.

And to answer another comment that they'll "believe it when they see the hardware" I'd have to say that's the most reasonable reaction to have when faced with these sorts of claims. I will say however that there are a lot of things we need to do on the software side before we reap the benefits of Infiniband and HPC. I'd love it if we could just plug the stuff in and see massive performance gains but that's not how it works.

One final note is that I'm just a game designer and while I have a passing knowledge of the technical stuff mostly through osmosis or lies on Wikipedia I'm no expert by any means and most of it seems like black magic to me. I'm sure many of you out there have more technical questions about this stuff but I'm the wrong guy to ask. I'll see if I can get someone else to comment but no promises.

---

Originally by: Neth'Rae

---

Originally by: Elektrea

---

Originally by: TheEndofTheWorld

---

1. CCP fixs lag for 600-1000 man in local.
2. Players start complaining about lag with 1000-2000 in local.

---

your a tard

---

No you are, this man speaks the truth..
Fleetbattles will just continue to grow in size as the population and alliances grow bigger.

Calling people on the forums "tards" for pointing out that doesn't change anything except that people will think you're the tard..

---

Originally by: Miss 10percent

---

Originally by: Neth'Rae

---

Originally by: Elektrea

---

Originally by: TheEndofTheWorld

---

1. CCP fixs lag for 600-1000 man in local.
2. Players start complaining about lag with 1000-2000 in local.

---

your a tard

---

No you are, this man speaks the truth..
Fleetbattles will just continue to grow in size as the population and alliances grow bigger.

Calling people on the forums "tards" for pointing out that doesn't change anything except that people will think you're the tard..

---

SO I guess you are alright with terribad lag while fighting in low-sec with 20 vs 20? The lag issues need to be addressed regardless of what limits players may push in deep 0.0.

---

Originally by: Jim McGregor

---

Edited by: Jim McGregor on 02/08/2008 17:19:57

Originally by: CCP Hammer

---

I'll see if I can get someone else to comment but no promises.

---

Moaaarrr hamsters!!

Thats my comment. Hope everybody is happy with it, I worked on it for quite some time.

---

Originally by: Ruze

---

Slow servers?
You mean, slow ISP's? Or are you saying that the servers CCP has, the ones that are the 'latest tech not even on the market yet' servers are slow?

---

Originally by: Neth'Rae

---

Originally by: Doc Fury

---

So... approximately how many lines of code have you guys "done on the software side" so far? In other words, how much EVE server code has already been rewritten for Infiniband?

I'm guessing this stuff is all still up on whiteboards and exists only in the realm of "possibilities" so far.

---

Your guessing doesn't help us a bit, go whine somewhere else or fly over to iceland and show them how to code their software..

---

Originally by: Doc Fury

---

Originally by: Neth'Rae

---

Originally by: Doc Fury

---

So... approximately how many lines of code have you guys "done on the software side" so far? In other words, how much EVE server code has already been rewritten for Infiniband?

I'm guessing this stuff is all still up on whiteboards and exists only in the realm of "possibilities" so far.

---

Your guessing doesn't help us a bit, go whine somewhere else or fly over to iceland and show them how to code their software..

---

CCP could not afford me.

I give you 4/10 for a troll attempt though.

---

Originally by: Neth'Rae

---

Originally by: Doc Fury

---

Originally by: Neth'Rae

---

Originally by: Doc Fury

---

So... approximately how many lines of code have you guys "done on the software side" so far? In other words, how much EVE server code has already been rewritten for Infiniband?

I'm guessing this stuff is all still up on whiteboards and exists only in the realm of "possibilities" so far.

---

Your guessing doesn't help us a bit, go whine somewhere else or fly over to iceland and show them how to code their software..

---

CCP could not afford me.

I give you 4/10 for a troll attempt though.

---

gb2wowforums...

---

Originally by: Neth'Rae

---

gb2wowforums...

---

Originally by: Complete Tart

---

Originally by: Neth'Rae

---

gb2wowforums...

---

Says the guy who wants to add wow-like Achievements to Eve

---

Originally by: Dantes Revenge

---

The hamsters are outdated and far to slow. We need rats.

---

Originally by: LaVista Vista

---

Originally by: Ruze

---

Or are you saying that the servers CCP has, the ones that are the 'latest tech not even on the market yet' servers are slow?

---

I will promise you, CCP hasn't got the latest technology on the market.

They haven't even got a real supercomputer yet. Their blade racks(They use IBM blades with Operton cores) are connected with high-end gigabit ethernet, which is slow in terms of super-computing, if not laughable.

The only "cool" thing they have, is their RAMDisk's. Those are some huge pieces of hardware with tons of RAM

---

Originally by: Benilopax

---

also thanks to CCP for coming to the meet and giving some free stuff to us.

---

Originally by: ImUrDestiny

---

Originally by: Benilopax

---

also thanks to CCP for coming to the meet and giving some free stuff to us.

---

They are giving out free T2 BPO's again???

didn't they learn anything from last time, i mean sjeez!!!

---

Originally by: Layla McScout

---

It doesn't matter how much hardware you toss at a shitty code. It will remain a shitty code, aka Python. Perharps calling Monty might remedy the situation a bit.

---

Originally by: Yaris San

---

As a new player to Eve I'm amazed by a few things.

1. The ability to check the market while in space.
2. The lack of an auto logout timer for long periods of afk.

It seems that tweaking these 2 points could greatly reduce lag. For example, you need level 5 trading to check the market while not docked and 15 minutes of afk gets you logged off.

I personally thing a lot of people are browsing the market while jumping and that has to cause a lot of server issues.

---

Originally by: CCP Hammer

---

I'd love it if we could just plug the stuff in and see massive performance gains but that's not how it works.

---

Originally by: Layla McScout

---

It doesn't matter how much hardware you toss at a shitty code. It will remain a shitty code, aka Python. Perharps calling Monty might remedy the situation a bit.

---

Originally by: Joakim Wasyl

---

Originally by: Layla McScout

---

It doesn't matter how much hardware you toss at a shitty code. It will remain a shitty code, aka Python. Perharps calling Monty might remedy the situation a bit.

---

Bit unfair, the code was designed / written 5 years ago when things were very different.

Given a clean sheet and know knowing the problem you can come up with a design that will scale nicely over multiple cores, processors etc and it's not particularly difficult.

CCP's problem though is how to move from what they have know, to any super new design without breaking everything. Any Software guys here will know this is far more difficult than simply starting from scratch.

---

Originally by: TheEndofTheWorld

---

1. CCP fixs lag for 600-1000 man in local.
2. Players start complaining about lag with 1000-2000 in local.

---

Originally by: Benilopax

---

So don't think that anyone who could fix bugs are being wasted on Ambulation it just isn't possible first because theose working on ambulation do not know how to bug fix and second each department concentrates on its own project.

---

Originally by: CCP Wrangler

---

EVE isn't designed to just look like a cold, dark and harsh world, it's designed to be a cold, dark and harsh world.

---

Originally by: Kerfira

---

Originally by: Benilopax

---

So don't think that anyone who could fix bugs are being wasted on Ambulation it just isn't possible first because theose working on ambulation do not know how to bug fix and second each department concentrates on its own project.

---

There is an easy solution to this....

Fire Ambulation department. Hire bug-hunters/game-developers!

There! Problem solved!

Not saying it SHOULD be done, but the 'excuse' that the resources spent on ambulation could not be used on more pressing issues is not based on reality!
Instead of hiding behind this 'excuse', CCP should just say "These are our priorities. This is how its going to be!"....

---

Originally by: bloomich

---

600v600 battles are not impossible. Its more a case of how long we have to wait before they become a reality. Is this reality going to be 5 years away? 5 months away? 50 years away?

---

Originally by: CCP Wrangler

---

EVE isn't designed to just look like a cold, dark and harsh world, it's designed to be a cold, dark and harsh world.

---

Originally by: Kerfira

---

Originally by: Benilopax

---

So don't think that anyone who could fix bugs are being wasted on Ambulation it just isn't possible first because theose working on ambulation do not know how to bug fix and second each department concentrates on its own project.

---

There is an easy solution to this....

Fire Ambulation department. Hire bug-hunters/game-developers!

There! Problem solved!

Not saying it SHOULD be done, but the 'excuse' that the resources spent on ambulation could not be used on more pressing issues is not based on reality!
Instead of hiding behind this 'excuse', CCP should just say "These are our priorities. This is how its going to be!"....

---

Originally by: Jimer Lins

---

Originally by: Kerfira

---

My post...

---

You're correct that those are CCP's priorities. You're incorrect in your assumption that software development works that way. You can't just move resources to another project and have it scale up.

In fact, doing so purely in the hopes of getting things done better or faster is one of the classic blunders of software development and almost always leads to failure.

---

Originally by: CCP Wrangler

---

EVE isn't designed to just look like a cold, dark and harsh world, it's designed to be a cold, dark and harsh world.

---

Originally by: HotSeat

---

200 vs 200, 600 vs 600 ..... Sorry CCP, 20vs20 doesn't work.

---

Originally by: Kerfira

---

Edited by: Kerfira on 03/08/2008 19:03:06

Originally by: bloomich

---

600v600 battles are not impossible. Its more a case of how long we have to wait before they become a reality. Is this reality going to be 5 years away? 5 months away? 50 years away?

---

The problem is the non-linear amount of load created for larger blobs.

Even if CCP doubled the amount of processing a single node oculd do, it might only mean a 40% increase in possible number of battle participants.

Let's say the load is quadratic. This means that X players generate X*X load. Lets take an example of 200 players.
This gives a load of 200*200 = 40000.
Double that and take the square root, and you get:
SQRT(2*40000) = 283 (rounded up)
So a doubling of the node capacity meant an increase of 41.5% in participant number.

Add to that the problem that less lag will just lead to bigger blobs until we lag out again.
What is needed is a change in game mechanics to discourage blobbing, and nobody has yet figured out what it is....

PS: Incidentally, if 200 people in a battle is what a node can handle today, it would take a 36-fold increase in node processing power to get to the 1200 people you use in your example

---

Originally by: Lord WarATron

---

For all you know, the server received 200 updates a tick from the 200 people, and then sends 1 update to each of these 200 people. So thats 400 updates per tick, not 200*200.

---

Originally by: Lord WarATron

---

Originally by: Kerfira

---

My post

---

Your math is wrong. 200*200? Your logic is wrong. "Lets not improve seating at this busy restaraunt because then more people will come"?

For all you know, the server received 200 updates a tick from the 200 people, and then sends 1 update to each of these 200 people. So thats 400 updates per tick, not 200*200. Its only 200*200 if you are some student in univeristy who thinks he knows it all without having spend 1 day working in the real world. Of course thats simplified because I do not know who CCP do it. But neither do you.

And the logic of not wanting ccp to improve to be able to handle more people? Seriously?

The only question is not technical ones. Trust me on this. The solution is if CCP have a plan to solve lag and how how they think it will be until they implement a solution they think is best to handle fleet battles.

---

Originally by: CCP Wrangler

---

EVE isn't designed to just look like a cold, dark and harsh world, it's designed to be a cold, dark and harsh world.

---

Originally by: Lord WarATron

---

Your math is wrong. 200*200? Your logic is wrong. "Lets not improve seating at this busy restaraunt because then more people will come"?

For all you know, the server received 200 updates a tick from the 200 people, and then sends 1 update to each of these 200 people. So thats 400 updates per tick, not 200*200. Its only 200*200 if you are some student in univeristy who thinks he knows it all without having spend 1 day working in the real world. Of course thats simplified because I do not know who CCP do it. But neither do you.

---

Originally by: Tarminic

---

Originally by: Lord WarATron

---

Your math is wrong. 200*200? Your logic is wrong. "Lets not improve seating at this busy restaraunt because then more people will come"?

For all you know, the server received 200 updates a tick from the 200 people, and then sends 1 update to each of these 200 people. So thats 400 updates per tick, not 200*200. Its only 200*200 if you are some student in univeristy who thinks he knows it all without having spend 1 day working in the real world. Of course thats simplified because I do not know who CCP do it. But neither do you.

---

Collision detection - it's usually O(N^2)/2 and there's no easy way to optimize it further.

---

Originally by: Tarminic

---

Originally by: Lord WarATron

---

Your math is wrong. 200*200? Your logic is wrong. "Lets not improve seating at this busy restaraunt because then more people will come"?

For all you know, the server received 200 updates a tick from the 200 people, and then sends 1 update to each of these 200 people. So thats 400 updates per tick, not 200*200. Its only 200*200 if you are some student in univeristy who thinks he knows it all without having spend 1 day working in the real world. Of course thats simplified because I do not know who CCP do it. But neither do you.

---

Collision detection - it's usually O(N^2)/2 and there's no easy way to optimize it further.

---

Originally by: Pan Crastus

---

That makes it very simple to:

a) implement for 1000s of objects on modern servers
b) parallelize

Reminder: phyics simulation done in Flash for 500 objects (press "5" key) with collision detection etc. ...

---

Originally by: Tarminic

---

Originally by: Pan Crastus

---

That makes it very simple to:

a) implement for 1000s of objects on modern servers
b) parallelize

Reminder: phyics simulation done in Flash for 500 objects (press "5" key) with collision detection etc. ...

---

Please, expand upon how you can optimize collision detection for consistently better performance than O(N^2)/2 in EVE's battles.

---

Quote:

---

Also, could you show me a similar application using 3D collision detection instead of 2D collision detection?

---

Originally by: LordCom

---

why they don't disconnect people who have 30 min of inactivity ? it's easy no ?

---

Originally by: Pan Crastus

---

You don't have to, because O(N^2)/2 is fine. Just because it doesn't have linear complexity, it doesn't mean it can't run in reasonable time with N=1000 or N=10000. Quicksort has O(N^2) in the worst case, go and test it with a worst case example and N=10000. And then please come back and argue that collision detection is what makes EVE slow because it's O(N^2)

---

Quote:

---

Quote:

---

Also, could you show me a similar application using 3D collision detection instead of 2D collision detection?

---

Why? Same complexity and the Flash example does not only collision detection, but also rigid body simulation (imagine EVE with 500 ships bumping each other.. nowdays you get desyncs with 5 ships).

---

Quote:

---

You know very well that we could get a 16x increase in "computation" performance easily if CCP cared to make their code multithreaded and buy a few servers with 16 cores, as everything in EVE is perfectly suited to parallelization. By your own arguments (O(N^2) etc.) that would mean 400x400 battles with the lag we get in 100x100 today, or 100x100 battles with today's lag with 25x25.

---

Quote:

---

Now for the networking issues that CCP undoubtedly has, I don't know enough to argue. All I can observe is that some people get huge latencies and others don't, so somehow I doubt that computation is the bottleneck always (or would the server treat entities on the same grid so unfairly regarding computation?). So my theory is that some of their uplinks get saturated quickly and others don't.

---

Originally by: Pan Crastus

---

Originally by: LordCom

---

why they don't disconnect people who have 30 min of inactivity ? it's easy no ?

---

What makes you think that they cause lag? They contribute the least to it ...

---

Originally by: Lord WarATron

---

Originally by: Tarminic

---

Originally by: Lord WarATron

---

Your math is wrong. 200*200? Your logic is wrong. "Lets not improve seating at this busy restaraunt because then more people will come"?

For all you know, the server received 200 updates a tick from the 200 people, and then sends 1 update to each of these 200 people. So thats 400 updates per tick, not 200*200. Its only 200*200 if you are some student in univeristy who thinks he knows it all without having spend 1 day working in the real world. Of course thats simplified because I do not know who CCP do it. But neither do you.

---

Collision detection - it's usually O(N^2)/2 and there's no easy way to optimize it further.

---

Does that make it impossible to play eve? Does that dictate how long it will be before CCP releases a solution to lag?

You can run collion detection with 400 objects on CPU power of a mordern day calculator with more accuracy than you get in eve currently, but of course, we are not comparing apples with apples there.

The issue is not that technical to be honest. it is not Computing issues that are relevant for university classes but not so relevant in real life, it is about finding out when a solution will be out and how far cpp are to implmenting one

Its like saying how long till we implement something that we think can break the speed of sound, rather than come up with various shockwave theories.

---

Originally by: Virida

---

The average mmorpg also uses half its cpu time running pathfinding, and thereunder, collision detection. Even if there is some extremely bad pathing in EVE's asteroid belts sometime, im happy with it as it is, if it make EVE stay different from EQ and WOW.

---

Originally by: Tarminic

---

And you know very well that unless you we have inside information there's no way to know what CCP's doing with their server code and how much they "care." Are you aware of any MMO with multithreaded server code (it would make an interesting case study, at least)?

---

Quote:

---

Quote:

---

Now for the networking issues that CCP undoubtedly has, I don't know enough to argue. All I can observe is that some people get huge latencies and others don't, so somehow I doubt that computation is the bottleneck always (or would the server treat entities on the same grid so unfairly regarding computation?). So my theory is that some of their uplinks get saturated quickly and others don't.

---

Extremely difficult to say given that we don't know much about their network. I can't do much other than shrug, as much as I enjoy theorycrafting.

---

Originally by: Tarminic

---

And you know very well that unless you we have inside information there's no way to know what CCP's doing with their server code and how much they "care."

---

Originally by: Pan Crastus

---

You must be kidding... If low-budget projects like vendetta online can write their server code in Erlang, it's safe to assume that any project leader who has been responsible for MMO development in the past 3-4 years and hasn't taken care to make it mulithreaded (or let's say "scaleable on SMP systems" because you don't need "threads" [kinda implies POSIX-style shared-data threading], Erlang-style lightweight processes will do), would probably get booted unless the producer is a moron.

---

Quote:

---

Even if they can't solve their bugs and bottlenecks, they could at least take care to make the lag more even ... It sucks to hear people playing fine on TS while half the fleet can do nothing for 15 minutes.

---

Originally by: HotSeat

---

200 vs 200, 600 vs 600 ..... Sorry CCP, 100 vs 100 doesn't work.

Have you put 100 people on a gate, and try to get anyone to jump in and not lag die?

This is why we are screaming about the speed nerf. In current code state, you can fly away at high speed well loading grid, most of the time (not always by any means)

Forget 600 vs 600 for now, just get current code working so that you don't appear in space until your client has loaded grid.

Don't claim "not possible", I've developed several of the first MMO's and would be glad to explain how.

---

Originally by: TheG2

---

Valve and TF2 are perfect examples of how difficult it is to code for Dual Core.

I'm a developer, I'm prepared to give CCP years of slack.

---

Quote:

---

Apache dies at about 4,000 parallel sessions. Yaws is still functioning at over 80,000 parallel connections.

---

Originally by: Zuteh

---

I have a question on this matter;
I have read around on this, and it seems like the problem is that the server nodes are not capable of scaling, like 1 grid, solarsystem, fight or <insert random speculation on CCPs server structure here> is locked at 1 blade/server.

So this leads me to; why did they chose python? Seriously, why the feck did they go with a no-no in multicore programming? Why did they not use Erlang? Erlang has been around forever (since 1987, that's forever in the pc-world), it is the shit when it comes to concurrent programming, it would be like made in heaven for EVE, you can scale whatever as much you want, you can just throw more hardware at the bottlenecks and it distribiutes itself-ish.

---

Originally by: Zuteh

---

So this leads me to; why did they chose python? Seriously, why the feck did they go with a no-no in multicore programming? Why did they not use Erlang? Erlang has been around forever (since 1987, that's forever in the pc-world), it is the shit when it comes to concurrent programming, it would be like made in heaven for EVE, you can scale whatever as much you want, you can just throw more hardware at the bottlenecks and it distribiutes itself-ish.

---

Originally by: Cpt Branko

---

...and throwing more programmers at a problem, particularly a iffy one involving rewriting present software which you have to understand, is likely to break more things then it fixes).

---

Originally by: Bartholomeus Crane

---

The problem is that without extensive changes to the code, Infiniband will hardly make any difference. Maybe some latency here, or a small speed-up there, but I don't expect more. To make use of this 'Next Generation I/O', you need change the code to hand off parts of the simulation. And that means concurrency, and lots of it. And that, ladies and gentlemen, a completely different can of worms.

---

Originally by: Pan Crastus

---

Originally by: Bartholomeus Crane

---

The problem is that without extensive changes to the code, Infiniband will hardly make any difference. Maybe some latency here, or a small speed-up there, but I don't expect more. To make use of this 'Next Generation I/O', you need change the code to hand off parts of the simulation. And that means concurrency, and lots of it. And that, ladies and gentlemen, a completely different can of worms.

---

Have they actually said yet how they'll use Infiniband? It might be that they are just replacing GigE with IPoIB for now, in which case it might actually have a noticeable effect with very little work involved (maybe the internal networking infrastructure is actually a huge bottleneck right now with ~30-50 systems per blade).

---

Originally by: Bartholomeus Crane

---

[After all, what use is it to plug in something like the RAMSAN if you can't get at the data fast enough?

---

Quote:

---

I rather think the underlying problem is with the indivisibility of the problem, or rather, that the code does not (sufficiently) make use of the concurrent nature of the hardware. Infiniband only makes sense when you're trying to address that. Otherwise, you'll have this super-fast memory disk, with a low latency, high throughput network, while still running large non-distributed code. It'll just stand there idling for all that money.

---

Quote:

---

People sometimes forget, HPC hardware requires HPC software to make use of it. All movies aside, having a Cray CX1 calculating single thread sequential tic-tac-toe is not very useful.

---

Quote:

---

The more I know about humans, the more I love animals.

---

Originally by: Grimpak

---

Edited by: Grimpak on 22/09/2008 14:21:10
[..]
so all in all, lag happens, in every game. And it's true that once you have lag-free 600vs600 battles, what will stop you from making 1000vs1000 battles, and lagging the game again?
[..]

---

Originally by: Saietor Blackgreen

---

Originally by: Zuteh

---

I have a question on this matter;
I have read around on this, and it seems like the problem is that the server nodes are not capable of scaling, like 1 grid, solarsystem, fight or <insert random speculation on CCPs server structure here> is locked at 1 blade/server.

So this leads me to; why did they chose python? Seriously, why the feck did they go with a no-no in multicore programming? Why did they not use Erlang? Erlang has been around forever (since 1987, that's forever in the pc-world), it is the shit when it comes to concurrent programming, it would be like made in heaven for EVE, you can scale whatever as much you want, you can just throw more hardware at the bottlenecks and it distribiutes itself-ish.

---

Read more, there's a relevant thread around. EVE system doesnt support multithreading inherently, due to the way the combat mechanics are processed. There are no lock mechanics, preventing things within one grid to desync against each other, and introducing those is a whole nightmare. Introducing this means recoding of whole EVE, creating EVE II.

Thats why they chose python - if you dont thread, you dont need threading tools.

Why was the system designed that way - thats a different question. Were there hopes for ongoiong increase in single core power back in 2000-2001, which failed, as the CPUs went multicore? Were they aiming for simpler system, as inroducing locks apperently complicates everything greatly? I have no idea, and I dont think it was ever stated by Devs. There may be many reasons.

---

Originally by: Zuteh

---

Edited by: Zuteh on 22/09/2008 14:49:32

Originally by: Grimpak

---

Edited by: Grimpak on 22/09/2008 14:21:10
[..]
so all in all, lag happens, in every game. And it's true that once you have lag-free 600vs600 battles, what will stop you from making 1000vs1000 battles, and lagging the game again?
[..]

---

I'm a beginner to Erlang, but wouldn't it be possible for a fleetfight to be done over multiple CPUs if the game was coded in Erlang? As I understand it, the resources are pooled in Erlang, the messages passed are extremely lightweight and you don't have to do things like manuall synchronization. The language is totally built upon concurrency.

---

Quote:

---

The more I know about humans, the more I love animals.

---

Originally by: Grimpak

---

Originally by: Zuteh

---

Edited by: Zuteh on 22/09/2008 14:49:32

Originally by: Grimpak

---

Edited by: Grimpak on 22/09/2008 14:21:10
[..]
so all in all, lag happens, in every game. And it's true that once you have lag-free 600vs600 battles, what will stop you from making 1000vs1000 battles, and lagging the game again?
[..]

---

I'm a beginner to Erlang, but wouldn't it be possible for a fleetfight to be done over multiple CPUs if the game was coded in Erlang? As I understand it, the resources are pooled in Erlang, the messages passed are extremely lightweight and you don't have to do things like manuall synchronization. The language is totally built upon concurrency.

---

don't know tbh. I'm no programmer

but, even if Erlang is that good, in the end the problem goes to point 1: total revamp from bottom up is very hard, expensive and dangerous, in a business POV.

---

Originally by: Elektrea

---

Originally by: TheEndofTheWorld

---

1. CCP fixs lag for 600-1000 man in local.
2. Players start complaining about lag with 1000-2000 in local.

---

your a tard

---

Originally by: Benilopax

---

So first, and I know this isn't about Lag but its worth mentioning. CCP has split into 3 departments: Outside Station (Spaceships and spacey things), Inside Station (The Market, Ambulation, refining etc etc), Bug fixing and Server (Obvious).

So don't think that anyone who could fix bugs are being wasted on Ambulation it just isn't possible first because theose working on ambulation do not know how to bug fix and second each department concentrates on its own project.


Discuss!

---

Quote:

---

Fire Abulation Team
Hire Members for other teams


The excuse "No one who could fix bugs is working on ambulation" cop out is simply that. Resources that are being used to focus on ambulation are being taken away from other departments.


Its simple fact that cant be debated away.

---

Originally by: Catharacta

---

My CNR runs on salvager tears.

---

Originally by: Kahega Amielden

---

Quote:

---

Fire Abulation Team
Hire Members for other teams


The excuse "No one who could fix bugs is working on ambulation" cop out is simply that. Resources that are being used to focus on ambulation are being taken away from other departments.


Its simple fact that cant be debated away.

---

If 10 developers can make a patch in 100 days, 100 developers can make a patch in 10 days.

Amirite?

---

Originally by: CCP Garthagk

---

While these forums may not give you everything that you want, they will usually let you post.

---

Originally by: Kahega Amielden

---

Quote:

---

Fire Abulation Team
Hire Members for other teams


The excuse "No one who could fix bugs is working on ambulation" cop out is simply that. Resources that are being used to focus on ambulation are being taken away from other departments.


Its simple fact that cant be debated away.

---

If 10 developers can make a patch in 100 days, 100 developers can make a patch in 10 days.

Amirite?

---

Quote:

---

Completely wrong actually. Experience in the software industry is the 100 developers will do it in 1000 days.
Go read some books on the subject, maybe

---

Originally by: Catharacta

---

My CNR runs on salvager tears.

---

Originally by: Kahega Amielden

---

Quote:

---

Fire Abulation Team
Hire Members for other teams


The excuse "No one who could fix bugs is working on ambulation" cop out is simply that. Resources that are being used to focus on ambulation are being taken away from other departments.


Its simple fact that cant be debated away.

---

If 10 developers can make a patch in 100 days, 100 developers can make a patch in 10 days.

Amirite?

---

Originally by: Huan Hunglong

---

More resources dedicated to any task normally (of course there are exceptions) reduces time taken on said task.

---

Originally by: Kropotkin

---

Edited by: Kropotkin on 23/09/2008 03:22:48

Originally by: Huan Hunglong

---

More resources dedicated to any task normally (of course there are exceptions) reduces time taken on said task.

---

In digging ditches, this is true.

In writing new software, not so much.

In fixing or improving old software, even less.

---

Originally by: Huan Hunglong

---

Originally by: Kropotkin

---

Edited by: Kropotkin on 23/09/2008 03:22:48

Originally by: Huan Hunglong

---

More resources dedicated to any task normally (of course there are exceptions) reduces time taken on said task.

---

In digging ditches, this is true.

In writing new software, not so much.

In fixing or improving old software, even less.

---

But as you state, it still applys.

---

Originally by: Yaris San

---

As a new player to Eve I'm amazed by a few things.

1. ...
2. The lack of an auto logout timer for long periods of afk.

It seems that tweaking these 2 points could greatly reduce lag. For example, you need level 5 trading to check the market while not docked and 15 minutes of afk gets you logged off.

---

Originally by: Kropotkin

---

http://top500.org/blog/2008/06/14
For what Roadrunner is built of, see page 15 of this PDF.

Well, ok, it's not Infiniband within each TriBlade.

But page 12 of that same PDF shows IB connecting the TriBlades.

Speculation about the relationship between Roadrunner and whatever IBM+CCP may be cooking up for the Future of Eve is ... left as an exercise for the reader. P-)

---

Originally by: Grimpak

---

3: With infiniband, the ability of actually splitting all the services and spreading them to several cores, will ease up the load per node. A single node will no longer process both fleetbattles, wallet updates, market transactions, chatlogs, combat logs, etc, etc, etc. That is good, since more cpu power = less lag.

---

Quote:

---

4: with load balancing, spreading a 500vs500 battle throughout even 2 nodes is bound to cause problems due to simple cpu-to-cpu desynch problems, software, and even the game itself, limitations.

---

Quote:

---

so all in all, lag happens, in every game. And it's true that once you have lag-free 600vs600 battles, what will stop you from making 1000vs1000 battles, and lagging the game again?

---

Quote:

---

CCP is trying to minimize lag, by upgrading the infrastructure (wich will allow for HPC and better upgradeablility in the future), and they can't do more unless they throw their sense of business out of the window,

---

Quote:

---

well this is what I got from that thread, but I also might be talking a big load of shit aswell.

---

Originally by: Pan Crastus

---

Quote:

---

well this is what I got from that thread, but I also might be talking a big load of shit aswell.

---

I see what you did there.

---

Quote:

---

The more I know about humans, the more I love animals.

---

Originally by: Varrakk

---

Any reasons why several blade servers cant support each other to run one system or a grid?

---

Originally by: Zanpt

---

Originally by: Yaris San

---

As a new player to Eve I'm amazed by a few things.

1. ...
2. The lack of an auto logout timer for long periods of afk.

It seems that tweaking these 2 points could greatly reduce lag. For example, you need level 5 trading to check the market while not docked and 15 minutes of afk gets you logged off.

---

Uh, being afk is not a load on the servers. You should open whatever your OS has for monitoring network traffic and actually look at the traffic your client generates while sitting in station or sitting in space where nothing is happening. Zip. Even while moving about there is very little traffic. Get a clue.

---

Originally by: Pan Crastus

---

Originally by: Zanpt

---

Originally by: Yaris San

---

As a new player to Eve I'm amazed by a few things.

1. ...
2. The lack of an auto logout timer for long periods of afk.

It seems that tweaking these 2 points could greatly reduce lag. For example, you need level 5 trading to check the market while not docked and 15 minutes of afk gets you logged off.

---

Uh, being afk is not a load on the servers. You should open whatever your OS has for monitoring network traffic and actually look at the traffic your client generates while sitting in station or sitting in space where nothing is happening. Zip. Even while moving about there is very little traffic. Get a clue.

---

I used to think so too, but consider jumping into a loaded system - what will your client load? A few 100 people's names and derived standings just for your local window... Now tell me that this doesn't affect e.g. the time it takes to jump to Jita.

Same with docking in Jita 4-4 ...

---

Originally by: Ratchman

---

As it has been observed by a few others, the more capacity you build into the system, the bigger the fleets will get by default.

CCP have been doing a good job of game development so far and although it hasn't evolved as quickly as I would have liked, I think there has been an acceptable level of adaptation. These things take time.

Of course, if you really wanted to avoid lag, you could always fly smaller fleets, but no-one's going to do that if they can outnumber the enemy.

Personally, I'll be happy if they get 200v200 fleets going without any lag. This should suffice for FW at least, and there is a natural limit for the number of people you can have in a fleet simultaneously (just try getting more than 50 people together for a FW fleet on any given day).

---

Originally by: Ratchman

---

As it has been observed by a few others, the more capacity you build into the system, the bigger the fleets will get by default.

CCP have been doing a good job of game development so far and although it hasn't evolved as quickly as I would have liked, I think there has been an acceptable level of adaptation. These things take time.

Of course, if you really wanted to avoid lag, you could always fly smaller fleets, but no-one's going to do that if they can outnumber the enemy.

Personally, I'll be happy if they get 200v200 fleets going without any lag. This should suffice for FW at least, and there is a natural limit for the number of people you can have in a fleet simultaneously (just try getting more than 50 people together for a FW fleet on any given day).

---

Originally by: Dr Slaughter

---

****BREAKING NEWS *****
CCP to hire forum posters in lag threads to work in their HPC cell.

The leader of the HPC cell stated today:

"We need 100,000 additional developers to help us complete our re-write of the core SOL services in Erlang and FORTH and to our suprise we discovered this fantastic resourec of experts in multi-threaded concurrent game design on our own forums! We still can't belive we got so lucky"

The register reports that the real and much more insidious reason for the recruitment drive is to test throwing live humans into lava tubes.

---

Originally by: Cheap Dude

---

This is an old topic brought up.. But will give my 2 cents anyways..

The big problem isn't the server capacity, its the gameplay thats the biggest problem. In 0.0 all is based on who has the biggest fleet and with the incomming speed nerfs this will even increase. In high-sec it's how the market works. The Jita problem can be resolved by introducing a max ammount of goods that can be sold within a system.

So instead of focusing on server capacity, CCP needs to focus on the gameplay to fix the lag.

---

Originally by: Bartholomeus Crane

---

Originally by: Cheap Dude

---

This is an old topic brought up.. But will give my 2 cents anyways..

The big problem isn't the server capacity, its the gameplay thats the biggest problem. In 0.0 all is based on who has the biggest fleet and with the incomming speed nerfs this will even increase. In high-sec it's how the market works. The Jita problem can be resolved by introducing a max ammount of goods that can be sold within a system.

So instead of focusing on server capacity, CCP needs to focus on the gameplay to fix the lag.

---

Surely you're not advocating some artificial limit, so I'll have to assume you mean some diminishing returns system to keep fleets small. Some good ideas about this have gone round, but all of them sounded a bit artificial to me. Perhaps you have some genius idea up your sleeve?

Either way, I do think that server capacity is a problem, the simple idea that a system is handled by a single CPU means that there's a hard limit on the maximum number of ships that can be handled efficiently. Moore's law isn't working anymore, concurrency is a solution, so CCP is right to focus on that.

Doesn't mean gameplay changes can't help though. They just look so artificial to me.

---

Quote:

---

advocating some artificial limit

---

Originally by: Cheap Dude

---

Hell no! I hate limits that can't be explained (ingame). But just look at the current 0.0 gameplay, its all based on bringing the most numbers. A big alliance always wins against smaller alliances, all they have to do is bring twice the numbers the enemy has and they win. The only thing thats stopping big alliances from winning within a small timeframe is the time it takes to bring down the posses.

About a market limit, yes.. please do so. In fact I find it strange all ships have a cargospace limit and stations don't. What would happen with Jita if they introduce a market limit, players would sell their items in another station/system (next system I think) which would resolve the massive players in one place (even in one station).

---