These forums have been archived and are now read-only.

The new forums are live and can be found at https://forums.eveonline.com/

EVE General Discussion

 
  • Topic is locked indefinitely.
 

CCP, I think today's events should be clear about this

Author
Krazynikomo
Perkone
Caldari State
#21 - 2013-01-27 19:12:04 UTC
Sentamon wrote:
A massive battle like that should involve weeks of buildup, not a spurt of the moment deal.


So, what you're saying is, CCP should ban goons?
Zagdul
Federal Navy Academy
Gallente Federation
#22 - 2013-01-27 19:14:06 UTC
Jaiimez Skor wrote:
True Sight wrote:
You are asking for something with a total lack of understanding of how these things work.

A reinforced node is basically a dedicated/larger node than usual, as soon as eve boots up after a maintenance systems start being dynamically loaded across the nodes (with special exceptions like Jita).

The loading screen you get when jumping from one system to another is often (but not always) a node transition too.

Once a node is already active, there's a piece of hardware handling all the stuff going on in that system, who is flying around where, who has who locked, missiles in flight, turrent mega-math, jamming, drones, reporting the locations of all this stuff to every other single player loaded on those same grids.

It's simply not feasible for a live seamless transition from one piece of hardware to another, you can do this with an Exchange or web-server quite easily, because the number of transactions and the response times are insignificant in comparison to an online game, where the millisecond response times are vital.

the only way to 'live reinforce' a node would be to basically either freeze the entire system (prevent people joining/leaving the system, stop all weapons fire, ship movement, velocity, active modules, active grids, grid sizes, other objects, all items in space etc etc etc), then transfer all that data over to one of the dedicated 'reinforced' nodes, then hit the pause button again whilst everyone continues.

The only other possibility would involve booting every single player out of the system (so... if the whole reason the fight started was due to a tackled super-fleet, said super fleet are now offline and safe and just won't log back in).

Time Dilation is the cure, its the only possibility in such spontaneous systems as your subscription and mine simply isn't enough for every single system in eve to be running on jita-level hardware for that one-percent chance that one day in the next few years it might suddenly host a huge spontaneous fleet battle.


Is true, however didn't CCP say that with TiDi that was capable, i'm sure at fanfest they said they do have the capability to put the TiDi at 0% and basically pause the game so it can be transferred mid fight to a reinforced node without any losses, i'm not sure if that still isnt' ready to happen and is still being worked on or what.


The benefit of TiDi is that your weapons hit their target (or miss if that's the case), your ship position is better calculated and everyone gets their turn.

It doesn't mean that the servers are capable of running the code to support almost 2000 players on a single grid with literally tens of thousands of drones and thousands of ships.

Dual Pane idea: Click!

CCP Please Implement

Zagdul
Federal Navy Academy
Gallente Federation
#23 - 2013-01-27 19:16:52 UTC
Sentamon wrote:
A massive battle like that should involve weeks of buildup, not a spurt of the moment deal.


This isn't possible and the game is advertised to be the exact opposite of this sentiment.

Spur of the moment battles are part of this game, some are just bigger than others and escalations like the one that happened last night are nearly impossible to predict.

Who would imagine that almost 300 alliances would be in a single system? Nobdoy, but expect more of this in the future. Not planning for this is naive.

Dual Pane idea: Click!

CCP Please Implement

Pohbis
Neo T.E.C.H.
#24 - 2013-01-27 19:26:50 UTC  |  Edited by: Pohbis
True Sight wrote:
I respond to this post with experience btw, I've been in fear of loosing my carrier pre-TD, with myself and a conga-line of over a 100caps sitting in a system we can't leave because of node-death, no idea if my ship will be alive, I can see it appears cloaked, my little remaining Isotopes used, this is far worse than TD.

This 'is' possible, but what you're talking about, or rather asking, is an entire re-write of a significant amount of code, perhaps the entire core engine because obviously as the programmers have been writing eve over the years, they've never coded towards the functionality of a pause button, let alone the services taking place to live-transfer mid-fight 2,000 players from one node to another 100% reliability
Well I was primarily responding to another poster who made it sound like such a solution wouldn't be preferable to what we have today.

What you're talking about is not really proof that it is possible thou.

The fact that the servers got so bogged down before TiDi that the commands simply piled up until the node died, leaving the impression that the node was "paused" and not in fact executing commands more or less at random, does not mean that it's actually possible to freeze an entire system in place and move it to another node, allowing the fight to pick up exactly where it was frozen.

However, if the client/server supports going to 10%, wouldn't it be fair to assume that it can support going to 0% TiDi?

In theory this should allow the client to stop sending commands completely but not disconnect. What would need to happen then, is for the server to clear the backlog and once done, being able to dump all info to a dedicated note all while keeping the client informed that it is in a "transfer" state and not disconnected. Once the transfer is done, the server would need to inform the client to pick up and become active again.

Now I am no tech guru, but it seems that we will never actually have a cluster that is capable of dynamic load balancing.

What we do have however, is the ability for a system to be taken offline and brought back up again on a dedicated node. The main issue being that the system in question is most likely run together with a lot of other systems on the same node, so being able to drop individual systems would be necessary. Or, worst case, pause every system on the node in order to bring it back up without the affected system. I'm sure people not involved in the fight would prefer a few minutes of client idling, to playing the rest of their session on a node that is hosting a 10% TiDi fight somewhere.

With TiDi and the ability to slow down the client in order for the server to catch up, the final piece missing is the ability to save the state of a system and everything in it, and bring up that dedicated node with the info already on it.

The most crucial part in terms of the player experience, is for it all to be possible while the client is still connected and idling, as to not require a re-log and a scenario where players are missing when the system is starting up again.
True Sight
Deep Freeze Industries
#25 - 2013-01-27 21:07:08 UTC
Pohbis wrote:
The fact that the servers got so bogged down before TiDi that the commands simply piled up until the node died, leaving the impression that the node was "paused" and not in fact executing commands more or less at random, does not mean that it's actually possible to freeze an entire system in place and move it to another node, allowing the fight to pick up exactly where it was frozen.

However, if the client/server supports going to 10%, wouldn't it be fair to assume that it can support going to 0% TiDi?

In theory this should allow the client to stop sending commands completely but not disconnect. What would need to happen then, is for the server to clear the backlog and once done, being able to dump all info to a dedicated note all while keeping the client informed that it is in a "transfer" state and not disconnected. Once the transfer is done, the server would need to inform the client to pick up and become active again.

Now I am no tech guru, but it seems that we will never actually have a cluster that is capable of dynamic load balancing.

What we do have however, is the ability for a system to be taken offline and brought back up again on a dedicated node. The main issue being that the system in question is most likely run together with a lot of other systems on the same node, so being able to drop individual systems would be necessary. Or, worst case, pause every system on the node in order to bring it back up without the affected system. I'm sure people not involved in the fight would prefer a few minutes of client idling, to playing the rest of their session on a node that is hosting a 10% TiDi fight somewhere.

With TiDi and the ability to slow down the client in order for the server to catch up, the final piece missing is the ability to save the state of a system and everything in it, and bring up that dedicated node with the info already on it.

The most crucial part in terms of the player experience, is for it all to be possible while the client is still connected and idling, as to not require a re-log and a scenario where players are missing when the system is starting up again.


Code wise, there is a very significant difference between slowing things down and stopping entirely. You're also missing out on dozens of different factors that all make this a much larger more complicated mess. you're asking some sort of server script that is capable of transferring from one (paused) active piece of hardware to another information that will be running in active RAM such as

- Players ships
- Their location
- Their vector
- Their speed
- Their ship fitting
- Their active modules
- Their current HP
- Who they have locked (including active modules mid-cycle)
- Who they are shooting (including guns mid-reloading)
- Active drones, the drones locations, their current orders, vector, velocity, who they are trying to shoot, their hp
- Grids, grid sizes, grid locations, players mid-warp, wrecks, wreck contents (normally destroyed on-restart), corpses, gang link bonuses

Along with this how do you tackle this paused transfer for people in neighboring systems?

- You can't let them jump into the system before everyone else has loaded, this would give them an advantage
- Making them wait could cause them to become stuck at a gate and become killed by a hostile force, thats unfair
- The paused players are at a disadvantage whilst other players are preparing and getting ready to move their support fleets into position, this could be the vital seconds needed to drop a dozen triage carriers onto a dying titan
Ohishi
Apocalypse Reign
#26 - 2013-01-27 21:19:59 UTC
True Sight wrote:
You are asking for something with a total lack of understanding of how these things work.

A reinforced node is basically a dedicated/larger node than usual, as soon as eve boots up after a maintenance systems start being dynamically loaded across the nodes (with special exceptions like Jita).

The loading screen you get when jumping from one system to another is often (but not always) a node transition too.

Once a node is already active, there's a piece of hardware handling all the stuff going on in that system, who is flying around where, who has who locked, missiles in flight, turrent mega-math, jamming, drones, reporting the locations of all this stuff to every other single player loaded on those same grids.

It's simply not feasible for a live seamless transition from one piece of hardware to another, you can do this with an Exchange or web-server quite easily, because the number of transactions and the response times are insignificant in comparison to an online game, where the millisecond response times are vital.

the only way to 'live reinforce' a node would be to basically either freeze the entire system (prevent people joining/leaving the system, stop all weapons fire, ship movement, velocity, active modules, active grids, grid sizes, other objects, all items in space etc etc etc), then transfer all that data over to one of the dedicated 'reinforced' nodes, then hit the pause button again whilst everyone continues.

The only other possibility would involve booting every single player out of the system (so... if the whole reason the fight started was due to a tackled super-fleet, said super fleet are now offline and safe and just won't log back in).

Time Dilation is the cure, its the only possibility in such spontaneous systems as your subscription and mine simply isn't enough for every single system in eve to be running on jita-level hardware for that one-percent chance that one day in the next few years it might suddenly host a huge spontaneous fleet battle.

COD pauses the game while it migrates hosts, after a certain point of TiDi maybe EVE should do the same thing.

Do not seek to follow in the footsteps of the wise. Seek what they sought.

True Sight
Deep Freeze Industries
#27 - 2013-01-27 21:42:25 UTC
Ohishi wrote:
COD pauses the game while it migrates hosts, after a certain point of TiDi maybe EVE should do the same thing.


Does COD have 2000 players in the same localised area? :) are players target locks, vectors, speed, active bullets maintained during the transition and preserved when the migration is done? :)
lanyaie
Nocturnal Romance
Cynosural Field Theory.
#28 - 2013-01-27 21:45:31 UTC
Cat Troll wrote:
[quote=Sairi Katelin]
P.S:
The player count was the same as what you see daily in Jita, so I don't think it would be that much computing power.


Because all the players in jita are pressing buttons and shooting and going around....no they are all docked unlike that fight

Spaceprincess

People who put passwords on char bazaar Eveboards are the worst.

Ohishi
Apocalypse Reign
#29 - 2013-01-27 21:53:23 UTC
True Sight wrote:
Ohishi wrote:
COD pauses the game while it migrates hosts, after a certain point of TiDi maybe EVE should do the same thing.


Does COD have 2000 players in the same localised area? :) are players target locks, vectors, speed, active bullets maintained during the transition and preserved when the migration is done? :)

Other than the 2000 people, yes. There is no reason that EVE shouldn't be able to do it.

Do not seek to follow in the footsteps of the wise. Seek what they sought.

Darvaleth Sigma
Imperial Security Hegemony
#30 - 2013-01-27 21:55:49 UTC
Have you heard of the Three body problem? When there are three objects it becomes very complicated to map out what their actions will be.

There were, what, 2800 bodies of varying mass, velocity, acceleration, and density, all in one system? Each with weaponry of varying damages, powered by capacitors of varying recharge rate and capacity, or by ammunition of varying damage to various defences, and of various speed and range? And every single ship with pilots of different skill levels and modules, creating thousands of %-changes that all impact on the combat? And THEN you throw in the fact that all of these are player-controlled, and that an internet signal needs to be sent out to all 2,800 players?

I'm surprised CCP isn't emblazoned across the skies, and that with their computing prowess they haven't already built a stargate.

Give a man a match and you warm him for a day.

Set a man on fire and he's warm for the rest of his life!

True Sight
Deep Freeze Industries
#31 - 2013-01-27 22:00:04 UTC  |  Edited by: True Sight
Ohishi wrote:
True Sight wrote:
Ohishi wrote:
COD pauses the game while it migrates hosts, after a certain point of TiDi maybe EVE should do the same thing.


Does COD have 2000 players in the same localised area? :) are players target locks, vectors, speed, active bullets maintained during the transition and preserved when the migration is done? :)

Other than the 2000 people, yes. There is no reason that EVE shouldn't be able to do it.


Problem is, the 2,000 is the key factor.
Ai Shun
#32 - 2013-01-27 22:09:42 UTC
Cat Troll wrote:
The player count was the same as what you see daily in Jita, so I don't think it would be that much computing power.


I'm not a cat. There is however a difference between sitting in Jita 4 and being in a slightly dented hull with a few carrier fulls of drones, missiles and other molten mayhem streaking across space around you.
Acac Sunflyier
The Ascended Academy
Space Potatoes Alliance
#33 - 2013-01-27 22:11:45 UTC
Cat Troll wrote:
Nodes need to auto-reinforce themselves when they are overloaded!


Thing is, they'd have to stop the fight anyway. to auto reinforce the node. The node has to be moved from one machine to another. This takes time to transfers all the relevant information.
Ai Shun
#34 - 2013-01-27 22:14:41 UTC  |  Edited by: Ai Shun
True Sight wrote:
the only way to 'live reinforce' a node would be to basically either freeze the entire system (prevent people joining/leaving the system, stop all weapons fire, ship movement, velocity, active modules, active grids, grid sizes, other objects, all items in space etc etc etc), then transfer all that data over to one of the dedicated 'reinforced' nodes, then hit the pause button again whilst everyone continues.


I'm not entirely convinced. With the way virtualisation is right now you can already have systems automatically migrate from host to host, effectively shifting it to a system with more processing power available. You can have dynamic ranges of resource allocation (memory I know you can, CPU I'm not 100% sure about but I can ask our hosting team).

It would of course depend on how they have things structured. If they're transferring only the game data across; rather than a slew of virtual servers then yes; there should be a freeze / etc. unless there is an arbiter system that can coalesce the two. But if they are simply migrating virtual servers (Even a volume of them) then it should be in the realms of feasibility with the way things are right now.

Acac Sunflyier wrote:
Cat Troll wrote:
Nodes need to auto-reinforce themselves when they are overloaded!


Thing is, they'd have to stop the fight anyway. to auto reinforce the node. The node has to be moved from one machine to another. This takes time to transfers all the relevant information.


NAS + associated benefits?
ACE McFACE
Dirt 'n' Glitter
Local Is Primary
#35 - 2013-01-27 22:17:16 UTC
Cat Troll wrote:

Dominiontrailerlolnope

The domination trailer is only 200 v 200 or something (plus that extra fleet at the end) Im pretty sure that wont break the server

Now, more than ever, we need a dislike button.

Vaal Erit
Science and Trade Institute
Caldari State
#36 - 2013-01-27 22:30:30 UTC
From 1 jump away staring at the system trying to load for 30+ minutes it was my understanding that the fight starrted and it got to 10% tidi very quickly and once the word got out that there were super cap kms to get onto everybody tried to get in and everything went to hell.

I feel like the people starting the fight and in it should get the lion's share of resources. I'd rather have the people in the battle at minimum 20% tidi and it being a lot harder to get in reinforcements until people die and they can be replaced. I could have used a "get me the hell outta here" button to get me to a different system because once you are stuck in that 10% tidi land, you are *stuck*
Xenuria
#37 - 2013-01-27 23:25:00 UTC
Ok here is a serious post.

I think that population caps for systems are a cop out.
I strongly feel that TiDi Should NOT be needed when a node is reinforced "properly".
I KNOW the technology exists, I also know it is super expensive.





My biggest problem is that even when a node is reinforced there is still horrible TiDi and lag. CCP has loads of money and they have server that make the servers of other MMOs look like potato lights. The problem still remains that for whatever reason or reasons CCP is not investing the money in better infrastructure.

Maybe I just don't have all the super secret info but this is how I and many other players see it.
Pohbis
Neo T.E.C.H.
#38 - 2013-01-27 23:55:25 UTC  |  Edited by: Pohbis
True Sight wrote:
Code wise, there is a very significant difference between slowing things down and stopping entirely. You're also missing out on dozens of different factors that all make this a much larger more complicated mess. you're asking some sort of server script that is capable of transferring from one (paused) active piece of hardware to another information that will be running in active RAM such as
No, the server is not paused. The client is. When the client is at 10% TiDi, the server is still running at full speed. That's kinda the point.

0% TiDi would allow the server to finish all calculations in the queue, while no new commands are being send from the connected clients.

Once the node has caught up, yes I am asking for it to dump the info. That's hardly rocket science. Any personal PC with a software RAMdisk can do this today.
Quote:
- Players ships
- Their location
- Their vector
- Their speed
- Their ship fitting
- Their active modules
- Their current HP
- Who they have locked (including active modules mid-cycle)
- Who they are shooting (including guns mid-reloading)
- Active drones, the drones locations, their current orders, vector, velocity, who they are trying to shoot, their hp
- Grids, grid sizes, grid locations, players mid-warp, wrecks, wreck contents (normally destroyed on-restart), corpses, gang link bonuses
Location, HP, speed, modules etc. are not an issue. The node stops accepting new commands and catches up before dumping the info. Now alignment might be an issue, but it's client side.Thou honestly, even if it required everyone not tackled to emergency warp when the re-inforced node is brought up, I think that's preferable over 10% TiDi slugfests. It's not like that wasn't the norm before TiDi. Except, you couldn't always count on the node being brought back up for the fight to continue. If it did, players would slowly trickle in one by one, skewing the fight, until the load reached levels were it would crash the node again.

Quote:
Along with this how do you tackle this paused transfer for people in neighboring systems?

- You can't let them jump into the system before everyone else has loaded, this would give them an advantage
- Making them wait could cause them to become stuck at a gate and become killed by a hostile force, thats unfair
- The paused players are at a disadvantage whilst other players are preparing and getting ready to move their support fleets into position, this could be the vital seconds needed to drop a dozen triage carriers onto a dying titan

- No of course not. Traffic control will prevent that. No different than today.
- Again, already happening today.
- They are already at a disadvantage in terms of outside support. They are fighting in 10% TiDi. If anything, quickly going from 10% TiDi to a re-inforced node should give the players in system the chance to drop a Titan before the enemy can get a response ready.
Lady Godwynn
The Scope
Gallente Federation
#39 - 2013-01-27 23:55:52 UTC
Cat Troll wrote:
Sairi Katelin wrote:
...Dude. With THAT MUCH STUFF in the system slapping each other around, I think CCP's coders are freaking awesomesauce to have kept the thing running as well as they did.

CCP server guys. You. Rule. Wish it'd have worked better, maybe someday we'll have the computing power available to have ludicrous spontaneous gigawars without headaches.

I'm not saying they didn't do a good job with the software and hardware.
A few years ago a battle half the size would crash EVE as a whole.
But until they can reinforce nodes automaticly one of the biggest things pushed in the advertisment can barely happen.

Dominiontrailerlolnope

P.S:
The player count was the same as what you see daily in Jita, so I don't think it would be that much computing power.


The limitation on players in Jita is something that I am interested in seeing increased as well. It can't be that resource
intensive when 80% of the people in system are docked up and just hitting the MSSQL server with their orders.

It happens every single weekend that you get the message that the system is overloaded and you're given the
choice of logging in next door.

This doesn't look good and players can't buy stuff!! :-)
Sergeant Acht Scultz
School of Applied Knowledge
Caldari State
#40 - 2013-01-28 00:05:10 UTC
lanyaie wrote:
Cat Troll wrote:
[quote=Sairi Katelin]
P.S:
The player count was the same as what you see daily in Jita, so I don't think it would be that much computing power.


Because all the players in jita are pressing buttons and shooting and going around....no they are all docked unlike that fight



Thousands of 0.1isk bot transactions, hundreds if not thousands of ship scans, modules activation/mission runners, guys shooting each other.
Jita is special

Lol

removed inappropriate ASCII art signature - CCP Eterne