Pages: [1] :: one page |
|
Author |
Thread Statistics | Show CCP posts - 2 post(s) |
JackofHearts
Together We Form Voltron
|
Posted - 2008.09.29 09:26:00 -
[1]
Edited by: JackofHearts on 29/09/2008 09:28:34 I have an idea to fix this problem...all trolling welcome...
break up eve solar systems to different servers or single cpu's...what I mean is perhaps one server can have 5 cpu's...
1 cpu for main system 1 cpu for battle sov area one 1 cpu for battle sov area two 1 cpu for battle sov area three 1 cpu for battle sov area four
Now I don't fully understand how jita runs cause it seems like it has it's own cpu or server. Is a node one server for other areas besides jita?
Anyways I think in order to take over a system you have to battle in 4 different sov staging points on there own independent cpu or server...so instead of of 500 vs 500 in a lagged out cluster fuk you have 100 vs 100 going on in 5 different areas. This could also make room for improvement on different types of ships and there attributes. Like lets face it some recon ships for example are not as fun to fly in big fleets as they are in smaller gangs. Perhaps this will make a difference in game play tactics as well as fun.
Basically you would have warp gates warpingor jumping from the same solar system they are fighting over to the sov battle areas where you have to kill a sov tower. I know the warp gates is not a very good idea but perhaps if there is a way to divide the fights up into 4 different staging areas that have to be simultaneous on 4 different cpu's or server boxes...
|
D desu
Demon Theory UNLeashed Legion
|
Posted - 2008.09.29 09:32:00 -
[2]
i dont think CCP would want to do that much coding to make something like that work. they would need to have there sever look at and identify "battle" areas with large player move all pro... you know what i cant be arsed typing... it would just be REALLY hard to do.
|
JackofHearts
Together We Form Voltron
|
Posted - 2008.09.29 09:34:00 -
[3]
Originally by: D desu i dont think CCP would want to do that much coding to make something like that work. they would need to have there sever look at and identify "battle" areas with large player move all pro... you know what i cant be arsed typing... it would just be REALLY hard to do.
I agree that it would be major and even change game play but if it works to fix the LAG?
maybe it would be worth it...
|
white kight
SwEaTy ArMpIT RaIDeRs
|
Posted - 2008.09.29 09:35:00 -
[4]
Wrong forum really as well.
......./me waves at Mitnal
Originally by: CCP Navigator
Thread Locked.
Please note that the General Discussion area is not....
Oh who am I kidding - Continue
|
JackofHearts
Together We Form Voltron
|
Posted - 2008.09.29 09:37:00 -
[5]
Originally by: white kight Wrong forum really as well.
......./me waves at Mitnal
this is for big alliances... sorry I thought this was the corp alliance discussion board...
|
Rakessh
Arachnea Phoenix Battalion Bionic Dawn
|
Posted - 2008.09.29 09:37:00 -
[6]
I'm sure CCP knows where the bottlenecks are from their telemetry data of their servers. I would assume they have already such crude work balancing as you suggest, but perhaps they need to work on parallelism of their code to allow their blade servers to scale up properly as load on single game solar systems explode (spiking local).
The problem seems from my player veiwpoint to be grid loading, suggesting that solar systems are indeed divided into process/threads by grid, and thus capable of running on a single cpu/core per grid.
Maybe the challenge is getting each grid to be cooperatively run by more than one cpu/core though, in which case that level of the design might need a total rewrite.
Or if that is too costly, just make due until faster processors can be put into the server farm :P
I think they go for the latter... ride it out ;)
PS: Total speculation on my part since I have absolutely no technical data on the software design :P
CEO Arachnea Phoenix Battalion |
Firkragg
Blue Labs Pandemic Legion
|
Posted - 2008.09.29 09:40:00 -
[7]
first of all what would happen is that each section in term would just get blobbed by the full amount of people possible.
Secondly your idea isnt really technically possible. If it was possible to split the action in the system between multiple nodes then instead of having your seperate section idea they can have the big fights like they currently do and just split the load. This is why they are developing the infiband technology and rewriting the code so this kinda thing is possible.
|
white kight
SwEaTy ArMpIT RaIDeRs
|
Posted - 2008.09.29 10:02:00 -
[8]
Originally by: JackofHearts
Originally by: white kight Wrong forum really as well.
......./me waves at Mitnal
this is for big alliances... sorry I thought this was the corp alliance discussion board...
It would be more at home in features and ideas form tbh
Originally by: CCP Navigator
Thread Locked.
Please note that the General Discussion area is not....
Oh who am I kidding - Continue
|
Dierdra Vaal
Caldari Eve University Ivy League
|
Posted - 2008.09.29 10:18:00 -
[9]
just adding more CPU doesnt solve the problems, it just moves the problem threshold.
Jita is an example: In the beginning, Jita would crap out at a few hundred people. Now, after many hardware and software upgrades, jita handles a few hundred people just fine. Problem over you'd think... but no! Since Jita can now handle a few hundred people, more people go into Jita. And once again we're stuck on sunday with a terribly laggy jita, because it has more than a thousand people in it.
Giving solarsystems a bigger capacity simply means alliances will put more people in it, resulting in the same lag.
Training Director :: EVE University
CSM Representative |
Virida
Mindstar Technology
|
Posted - 2008.09.29 10:20:00 -
[10]
Ok, lets assume we got a hypothetical CPU cluster running several cpu's. One solar system running on, say 4 cpu's in a server box.
What is needed?
Well, i do not imagine it possible to run a combat in 2 cpu's at same time, so all combat need to be done in 1 single cpu. And warping around in the system probably means you get internal zoning logic in the cluster, who could take a whole cpu to track, along with multicore bookkeeping.
A combat is monolithic in nature, 1 affect 2 affects 3 affects 4, etc. It are hard to divide a dynamic system of hundreds of independent entities in realtime combat into seperate entities for multicore processing. Well, i am not an experienced programmer, and dont really know multicore at all. But a lot can probably be done if you have direct memory access, or better memory access between cpu's, even in traditional blade servers.
|
|
Mang0o
Caldari The Priory
|
Posted - 2008.09.29 19:43:00 -
[11]
i think its better if we all just quit eve and get a real life |
Vogue
Gods Killboard Pwns All-But Does Not Pay The Bills
|
Posted - 2008.09.29 19:46:00 -
[12]
You have no idea how powerful the smite of the metaphysical is |
Anna Valerios
Amarr Rage and Terror Against ALL Authorities
|
Posted - 2008.09.29 19:50:00 -
[13]
God shut the **** up, no one gives a flying shit about your terrible idea. Go run some missions or something you moron.
|
Sapapaya
Caldari Division Solaris
|
Posted - 2008.09.29 19:57:00 -
[14]
What do you think about a cpu for your brain?
|
Vogue
Gods Killboard Pwns All-But Does Not Pay The Bills
|
Posted - 2008.09.29 20:03:00 -
[15]
Uh huh...
Madness is something rare in individualds - but in groups, parties, peoples it is the rule.
Btw the word is today is: OINK.
|
NAT Mav
TribalWar Inc
|
Posted - 2008.09.29 20:30:00 -
[16]
The existing limitations only allow a maximum of 1 CPU to be allocated per solar system, as is the case with Jita. In most cases, however, you have multiple solar systems assigned to a single CPU (node). It's like trying to run an older Windows application on a Dual-Core PC. The application doesn't know how to take advantage of the multiple CPUs, so you don't see a huge increase in performance over a single-core PC with similar specs while running just that one application.
Whether or not Stackless IO has removed this limitation, I do not know.
|
|
CCP Mitnal
C C P
|
Posted - 2008.09.30 01:52:00 -
[17]
Moved to Game Development. |
|
Trellish
Ten Ton Hammer The Red Skull
|
Posted - 2008.09.30 05:53:00 -
[18]
Hmmmm...
Seems like you make a mighty big presumption here... i.e. that the bottleneck is the CPU.
That seems a bit unlikely to me given the nature of this game. |
Silence iKillYouu
SniggWaffe
|
Posted - 2008.09.30 07:11:00 -
[19]
All lies. The only word that should interest you is NIPPLE.
Nipples R Us, that is all.
|
Hungo
Minmatar Pilots Of Honour Axiom Empire
|
Posted - 2008.09.30 10:15:00 -
[20]
Originally by: JackofHearts Edited by: JackofHearts on 29/09/2008 09:44:10 I have an idea to fix this problem...all trolling welcome...
break up eve solar systems to different servers or single cpu's...what I mean is perhaps one server can have 5 cpu's...
1 cpu for main system 1 cpu for battle sov area one 1 cpu for battle sov area two 1 cpu for battle sov area three 1 cpu for battle sov area four
Now I don't fully understand how jita runs cause it seems like it has it's own cpu or server. Is a node one server for other areas besides jita?
Anyways I think in order to take over a system you have to battle in 4 different sov staging points on there own independent cpu or server...so instead of of 500 vs 500 in a lagged out cluster fuk you have 100 vs 100 going on in 5 different areas. This could also make room for improvement on different types of ships and there attributes. Like lets face it some recon ships for example are not as fun to fly in big fleets as they are in smaller gangs. Perhaps this will make a difference in game play tactics as well as fun.
Basically you would have warp gates warping or jumping(to different server) from the same solar system they are fighting over to the sov battle areas where you have to kill a sov tower. I know the warp gates is not a very good idea but perhaps if there is a way to divide the fights up into 4 different staging areas that have to be simultaneous on 4 different cpu's or server boxes...
Hi Zyphentis
|
|
|
CCP Casqade
|
Posted - 2008.09.30 10:49:00 -
[21]
There are multiple threads about the server architecture and the bottlenecks on the forum. We are always looking into making things better, StacklessIO is a result of 2 years of work. |
|
|
|
|
Pages: [1] :: one page |
First page | Previous page | Next page | Last page |