These forums have been archived and are now read-only.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.
Previous page12
 

What is New Eden's capacity?

Author
James 315
Experimental Fun Times Corp RELOADED
CODE.
#21 - 2012-08-21 22:24:25 UTC
The capacity is functionally infinite. The more players they get, the more subscriptions, the more money, the more server upgrades... The income they get will always outstrip the extra demand on the server.

The only place where they get into trouble is when too many people appear unexpectedly at the same location. With warning, they can reinforce the node or even build a separate node for a system (like Jita). But as we used to see in nullsec fleet fights, and even during the Burn Jita event, capacity is an issue if they don't have time to prepare.
Steve Ronuken
Fuzzwork Enterprises
Vote Steve Ronuken for CSM
#22 - 2012-08-21 22:30:46 UTC
Tiger Would wrote:
Denidil wrote:
Tiger Would wrote:
Vera Algaert wrote:
Reicine Ceer wrote:
By that, I mean, what is the maximum number of players the server could handle before failing? I'm just curious.... I wonder how it compares to a single server in WoW? Whether TiDi would need to be implemented at every gate if the server reached maximum capacity? Or would CCP end up just getting a beefier server?

...these are the things that keep me awake at night.

TQ isn't one server but a whole farm of them - you have the proxy servers which handle the connections to the clients, you have the sol servers each of which hosts one or more solar systems and you have the database servers that everything talks to.

For some outdated numbers see:
http://wiki.eveonline.com/en/wiki/Tranquility#Current_capability_01-13-2011_from_CCP_Yokai



But are those nodes physical?

Because, I dont know, I t could be a virtualized environment, it would make sense in a load balancing energy usage kind of way. ( But also storage and bandwidth between DB and nodes (CUCS))


I could easily be wrong though.


they're not virtualized


If so, it might not be a bad idea to start thinking about that.....
VMWare has very nifty options in CUCS, QOS mechanics on IO for example....

Infiniband between physical and virtual nodes (traffic wise, maybe not needed, but latencies drop conciderably)
Huge bandwidth between storage and nodes

Energy usage drops concidrable if you let DRS also look at loads and start migrating nodes when only 20k people are online and switch on physical nodes and migrate again when the counter goes up.

Also CPU resource pooling could overcome the single core (ok now they sidebanded some traffic to a second core) limitations for quite a bit

And it is where IT is headed, para virtualization has reached limitations, binary translation can still grow, but hardware virtualization is just around the corner and partially already supported on some CPU's. (partially run code on CPU instead of hypervisor level)

Not to mention a much much lower COS (less physycal but beavier servers) less rack space, less cooling capacity needed, less UPS capacity, much lower power consumption etc


and so on.....


ah well.....it works like it is....


Actually, they are virtualised at the node level (though I can't find the reference. Dev blog or fanfest presentation. Possibly a Tidi one)

The nodes can be moved around as needed. Though they can't, yet, split systems from heavily loaded nodes.

Woo! CSM XI!

Fuzzwork Enterprises

Twitter: @fuzzysteve on Twitter

Tiger Would
Doomheim
#23 - 2012-08-21 22:34:23 UTC
Steve Ronuken wrote:


Actually, they are virtualised at the node level (though I can't find the reference. Dev blog or fanfest presentation. Possibly a Tidi one)

The nodes can be moved around as needed. Though they can't, yet, split systems from heavily loaded nodes.


makes sense, i did not want to doubt the nay sayer without knowing so left it in the middle

Once you think you have it all, you have actually become ignorant towards everything else.

T. Would

SmilingVagrant
Doomheim
#24 - 2012-08-21 22:45:13 UTC
I require clarification: number of players or girth of players?
Jett0
Ministry of War
Amarr Empire
#25 - 2012-08-21 23:53:05 UTC
Considering Dust will run on TQ, I imagine the servers are very capable.

Occasionally plays sober

Barakach
Caldari Provisions
Caldari State
#26 - 2012-08-22 02:44:14 UTC
I'm curious how they have their processing done. Modern CPUs are thread-strong, but I wonder how much critical code paths would hurt multi-threading.

One thing I've leaved about writing scalable code is learning when you don't need something "perfect". Many times I've found that preserving order or whatever is "nice", but not a requirement. Little things like this have allowed me to write much more scalable code.

It's not a "Race Condition", it's a "feature"!
Dennis Gregs
The Scope
Gallente Federation
#27 - 2012-08-22 03:22:31 UTC
Roll Sizzle Beef wrote:

An average peek for a WoW server is like 750.


The server capacity (amount of logged-in players needed to reach a "full" status with a queue) was never that low, and they increased the capacity with both Burning Crusade and Wrath of the Lich King. A WoW 'realm' currently can definitely carry 3500 people, if not more. I've heard it used to be a lot more but they reduced it due to issues, but it's all speculation.

However, due to the fact that a WoW 'realm' is a significantly smaller server cluster than what EVE's is, if you stockpile even only 500 people in the same place you start getting issues, and 500 might be generous.
Pipa Porto
#28 - 2012-08-22 04:44:40 UTC
Blane Xero wrote:
Roll Sizzle Beef wrote:
Current record recorded http://games.chruker.dk/eve_online/server_status.php
was 63,165 at once.

An average peek for a WoW server is like 750.

Perhaps for a low to mid population Realm.

After doing some looking into it, EU High pop realms average out to about 2k~ combined, with peaks of 4k~.


[Sampled with info from Argent Dawn EU.]



That's pretty cool. Some of EVE's battles have been bigger than WOW Servers.

EvE: Everyone vs Everyone

-RubyPorto

Gilbaron
The Scope
Gallente Federation
#29 - 2012-08-22 05:16:18 UTC
tq can probably manage more than double the current player activities before devs get nervous about server performance. throwing large amounts of new hardware on the problem will take the pressure for quite some time.

excluded are large numbers of players in a specific location, or even worse, these players switching locations (grids or worse, systems). These problems could possibly be solved by working on the code itself or underlying mechanics, as mentioned by veritas at fanfest, he was talking about 'brain in a box', a new way of calculating how a players actions are influenced by skills, implants and projected effects. right now this happens on the same server as everything else, in a (far) future eve that would be done separate from combat
Herr Hammer Draken
#30 - 2012-08-22 05:24:10 UTC  |  Edited by: Herr Hammer Draken
I am going to say we are nearly at capacity right now. I say that because on a daily basis Jita trips the TiDi feature already.

Now what will happen to EVE if we get say 100,000 accounts active on EVE at once? Well at least 40,000 of them will be in Jita all at the same time. Because it is what people will do go to Jita. Everybody goes to Jita. What wil that be like? I guess TiDi can slow it down enough that 1 second of game time will take 1 hour of real time to play. So a trip from undock to the exit gate out of the system takes what about 90 seconds of game time hey only 90 hours of real time after TiDi kicks in to handle the mad rush of players in Jita.

So yea we can take as many people as we want in EVE just need to be online forever when you go to Jita.

Herr Hammer Draken "The Amarr Prophet"

Asuka Solo
I N E X T R E M I S
Tactical Narcotics Team
#31 - 2012-08-22 06:34:33 UTC
Jimmy Gunsmythe wrote:
Take the maximum number of allowed people in a system, say Jita and multiply that by the number of systems in game. That's your maximum amount of people allowed on one server.


Thats quite a few million players....

Eve is about Capital ships, WiS, Boobs, PI and Isk!

Radius Prime
Tax Evading Ass.
#32 - 2012-08-22 08:56:04 UTC  |  Edited by: Radius Prime
Tau Cabalander wrote:
TQ Level Up

CCP Yokai wrote:
TQ Tech Details: (Not the whole system, just what runs TQ)

Servers
64 x IBM HS21
2x Dual Core 3.33GHz CPU's
32GB of RAM Each
1x72GB HDD Each

2 x IBM X3850 M2's
2x Six Core 2.66GHz
128GB of RAM
4 x 146GB HDD

Cores
- 280 total Cores
- ~1 THz

RAM
- 2.3TB of Total RAM

Storage
- 4.8TB of Local Storage
- 2TB of SSD SAN
- 256GB of RAM SAN

Network
- Gigabit Ethernet
- 4Gb/s Fiber Channel



That's pretty sad. No wonder we get lag wars. Upgrade already CCP. And let us fight serenity .

Reopen the EVE gate so we can invade Serenity. Goons can go first.

Steve Ronuken
Fuzzwork Enterprises
Vote Steve Ronuken for CSM
#33 - 2012-08-22 09:09:53 UTC
Radius Prime wrote:
Tau Cabalander wrote:
TQ Level Up

CCP Yokai wrote:
TQ Tech Details: (Not the whole system, just what runs TQ)

Servers
64 x IBM HS21
2x Dual Core 3.33GHz CPU's
32GB of RAM Each
1x72GB HDD Each

2 x IBM X3850 M2's
2x Six Core 2.66GHz
128GB of RAM
4 x 146GB HDD

Cores
- 280 total Cores
- ~1 THz

RAM
- 2.3TB of Total RAM

Storage
- 4.8TB of Local Storage
- 2TB of SSD SAN
- 256GB of RAM SAN

Network
- Gigabit Ethernet
- 4Gb/s Fiber Channel



That's pretty sad. No wonder we get lag wars. Upgrade already CCP. And let us fight serenity .


Check the link. 2 years ago.

Woo! CSM XI!

Fuzzwork Enterprises

Twitter: @fuzzysteve on Twitter

Solstice Project
Sebiestor Tribe
Minmatar Republic
#34 - 2012-08-22 16:31:14 UTC
Steve Ronuken wrote:
The nodes can be moved around as needed. Though they can't, yet, split systems from heavily loaded nodes.


This is only true for non-realtime applications.
They can't move around solars systems seamlessly yet.
Solstice Project
Sebiestor Tribe
Minmatar Republic
#35 - 2012-08-22 16:33:18 UTC  |  Edited by: Solstice Project
Herr Hammer Draken wrote:
I am going to say we are nearly at capacity right now. I say that because on a daily basis Jita trips the TiDi feature already.

Now what will happen to EVE if we get say 100,000 accounts active on EVE at once? Well at least 40,000 of them will be in Jita all at the same time. Because it is what people will do go to Jita. Everybody goes to Jita. What wil that be like? I guess TiDi can slow it down enough that 1 second of game time will take 1 hour of real time to play. So a trip from undock to the exit gate out of the system takes what about 90 seconds of game time hey only 90 hours of real time after TiDi kicks in to handle the mad rush of players in Jita.

So yea we can take as many people as we want in EVE just need to be online forever when you go to Jita.


That's pretty much wrong because Jita has it's own (probably stronger) node and the more people want to be in Jita
the stronger they'll make it.

Jita is in no way ever a base to measure the capacity of TQ, unless they put up a node for every system
and every node is as strong as the jita node.
Jim Era
#36 - 2012-08-22 16:39:48 UTC
Herr Hammer Draken wrote:
I am going to say we are nearly at capacity right now. I say that because on a daily basis Jita trips the TiDi feature already.

Now what will happen to EVE if we get say 100,000 accounts active on EVE at once? Well at least 40,000 of them will be in Jita all at the same time. Because it is what people will do go to Jita. Everybody goes to Jita. What wil that be like? I guess TiDi can slow it down enough that 1 second of game time will take 1 hour of real time to play. So a trip from undock to the exit gate out of the system takes what about 90 seconds of game time hey only 90 hours of real time after TiDi kicks in to handle the mad rush of players in Jita.

So yea we can take as many people as we want in EVE just need to be online forever when you go to Jita.



I never went to jita.
Your post is invalid.

Wat™

Boxless
Deep Core Mining Inc.
Caldari State
#37 - 2012-08-22 18:56:30 UTC
Barakach wrote:
I'm curious how they have their processing done. Modern CPUs are thread-strong, but I wonder how much critical code paths would hurt multi-threading.

One thing I've leaved about writing scalable code is learning when you don't need something "perfect". Many times I've found that preserving order or whatever is "nice", but not a requirement. Little things like this have allowed me to write much more scalable code.

It's not a "Race Condition", it's a "feature"!



EVE was coded in such a way that makes it so that a node(1-50 solarsystems coupled) can not be split among multiple threads easily(meaning that on a normal node a system can use at most 1 entire core).

CCP does have a handful of nodes that has the capability(they are considerable expensive) through virtualizing multiple cores to act as one, those nodes are used for trade hubs and fleet fights.

From what I understand CCP is working on slowly decoupling tasks in order to allow multi-threaded server architecture in the future. We are not to expect that anytime soon.

 FREE Iamien !!!!!

Methesda
Sebiestor Tribe
Minmatar Republic
#38 - 2012-08-22 21:35:51 UTC
Eve's capacity is one less than we have now. I think you should prostrate yourself infront of me, so that I may kill you and save this universe.

Thank you.

Eve is about the journey.  If you are so focused on making money, that you insist on having the tools to make it be made as autonomous and easy as possible, then you are never going to have as much fun as I will.

Ensign X
#39 - 2012-08-22 21:44:22 UTC
Considering how terrible this game is to play when you get a few hundred people in the same system, let's hope the population never spikes anywhere close to whatever it's theoretical capacity is.
Previous page12