| Pages: [1] :: one page |
| Author |
Thread Statistics | Show CCP posts - 0 post(s) |
Delerium Primus
M. Corp
M. PIRE
   |
Posted - 2007.09.02 16:14:00 -
[1]
Edited by: Delerium Primus on 02/09/2007 16:23:13
Just struck me that much of the processing power of the eve tranquility cluster is non combat related and thus not response time intensive - eg. market/contracts etc dont need an immediate response and could be done every 10 seconds or so. So with this in mind, would it be possible to implement something like the 'folding at home initiative' to lighten the load on the cluster to boost performance?
For those unfamiliar with this initiative, it invloves utilising spare cpu power of clients machines to do processing thus giving you thousands/millions of computers to complete simple cpu processes without relying on one main computer.
CCP constantly upgrade their hardware to try and upgrade the power of their machines but how far can this go? limitations are technology advances and, lets be honest, cash. Creating a scheme like this may be just what eve needs in order to divert resources to gameplay and leave the backend processes like markets etc to be done via the spider of clients that eve has.
*EDIT*
Just to add - I know a lot of you will say "Eve is already massivley CPU intensive" and the answer is yes it is but that is being worked on with a new GFX engine that will hopefully divert all the processes currently done by cpu to your GFX card leaving spare CPU power. Even with a CPU intensive client, my Eve runs only at 50% CPU per client leaving plenty of spare cpu for spidering.
Your thoughts?
Del
|
Hanshin Kool
Amarr
Royal Amarr Institute
|
Posted - 2007.09.02 16:23:00 -
[2]
I don't think it would work. Folding@home and SETI and the other distributed projects use the fact that the math they are trying to do can be broken up into tiny chunks which can be done a piece at a time by other machines. Calculating the damage your ship is going to take from that incoming missile isn't something that can wait for Jeb Thornbush's computer in Little Rock, Arkansas to compute while your ship's position has alrerady been updated by Enda Flatbottom's computer in London, England. It needs to all be done in a certain order.
|
Delerium Primus
M. Corp
M. PIRE
|
Posted - 2007.09.02 16:25:00 -
[3]
 Originally by: Hanshin Kool
I don't think it would work. Folding@home and SETI and the other distributed projects use the fact that the math they are trying to do can be broken up into tiny chunks which can be done a piece at a time by other machines. Calculating the damage your ship is going to take from that incoming missile isn't something that can wait for Jeb Thornbush's computer in Little Rock, Arkansas to compute while your ship's position has alrerady been updated by Enda Flatbottom's computer in London, England. It needs to all be done in a certain order.
I agree with you but Im not suggesting this be used for combat of any sort - that is time specific. There are a lot of processes that aren't. To name a few:
POS Fuel Hourly Ticks
NPC Buy/Sell Orders
Market Updates
Map Statistic Updates
|
C0CK KN0CKER
Mail Order Russian Brides
|
Posted - 2007.09.02 18:34:00 -
[4]
If CCP do implement a method to improve CPU and reduce lag, surely it would reduce the number of petitions due to lag, which would save the CPU needed to create the auto-reponse of 'Our server logs show nothing to support your case, we hope you recover quickly'
Either that, or extra CPU may be enough to enable server logging, which can show what's actually happening (or not as the case may be) in game.
Who knows, extra power may even re-introduce the long awaited 'fleet battles', ones you can actually participate in, and view - instead of having a blank screen and then waking up in a station.
-----------------
Mailorder Brides just for you. Free isk if ordered before May |
Delerium Primus
M. Corp
M. PIRE
|
Posted - 2007.09.02 19:58:00 -
[5]
Hmmm I was sure that people might go for this as this would be a very effective way to reduce server load. 80+ reads and only 2 comments?
Del
|
Ellaine TashMurkon
MetaForge
Ekliptika
|
Posted - 2007.09.02 21:26:00 -
[6]
Edited by: Ellaine TashMurkon on 02/09/2007 21:27:08
It brings risk of really bad hacking attacks.
It can't work. There are no CPU super heavy operations with light data packages, no brute forcing viruses and alien communication. Only tons of very simple operation. Drone moves with speed S, in direction V, calculate new position after 2 seconds etc. about 32 bytes (assuming everything is in 8 byte float), about 3 multiplications and 3 additions, calculations near to none. Not worth sending, checksuming etc.
In system You're talking about You have a package like 20k (some part of signal from space), and millions of operations to do with it.
|
Drizit
Amarr
Lonely out here
Black Sun Alliance
|
Posted - 2007.09.03 02:13:00 -
[7]
One of my corp has a dongle linked via GPRS on his laptop. O2 cap his dl limit to 3 gig which means spidering will reduce his game time.
I really think that will go down a treat with him and many others who are capped.
--
|
Ivanna Humplot
|
Posted - 2007.09.03 02:40:00 -
[8]
I dont think he is talking here about flight, combat or anything like that. He is talking about the back end processes that you dont see. Sure it could be prone to hacks but can't the eve client? I definitley see some milage in this.
/signed
Ivanna
|
Ciara Daag
|
Posted - 2007.09.03 06:52:00 -
[9]
Originally by: Delerium Primus
Originally by: Hanshin Kool
I don't think it would work. Folding@home and SETI and the other distributed projects use the fact that the math they are trying to do can be broken up into tiny chunks which can be done a piece at a time by other machines. Calculating the damage your ship is going to take from that incoming missile isn't something that can wait for Jeb Thornbush's computer in Little Rock, Arkansas to compute while your ship's position has alrerady been updated by Enda Flatbottom's computer in London, England. It needs to all be done in a certain order.
I agree with you but Im not suggesting this be used for combat of any sort - that is time specific. There are a lot of processes that aren't. To name a few:
POS Fuel Hourly Ticks
NPC Buy/Sell Orders
Market Updates
Map Statistic Updates
This would be a very very bad idea. If you use peoples computers to do game processing,then people will exploit that to cheat. It would open huge security flaws. The reason they can do it with folding at home and seti @home is because the security isn't an issue. If someone gets a hit,they recheck that data block on their own computer. You can cheat all you want and upload fake hits to their server,but at the end of the day,if they recheck it and it isn't legit,all you did was got yourself banned from the project.
|
Ciara Daag
|
Posted - 2007.09.03 07:11:00 -
[10]
Edited by: Ciara Daag on 03/09/2007 07:15:07
Originally by: Ivanna Humplot
I dont think he is talking here about flight, combat or anything like that. He is talking about the back end processes that you dont see. Sure it could be prone to hacks but can't the eve client? I definitley see some milage in this.
/signed
Ivanna
I don't know about eve specifically,but whats usually done client side with hacks are things like speed or movement hacks,or gaining access to information that your not supposed to have,like for instance a monster just spawned in your zone. For instance,in Final Fantasy Eleven there was a hack that would let you see when a monster spawned and teleport to its location. When you actually give people access to back end data,you open up a whole new set of hacks. Imagine a hack that allowed you to change market prices that other players just put up,or something that tapped into the isk transfer processing and let you make a large transfer from someones wallet to your own. Imagine getting into the pos fuel processing and making your enemies pos run out of fuel so you can attack it. Imagine getting into the contract system and making them appear that your character had just purchased one.Its true that each one of those issues could be resolved in various ways but the problem is that your playing a giant game of whack a mole. You cant fix most of the problems until they are found and as soon as you fix one,someone else finds a new one to exploit.
You have to be secure by design,not as an after thought. Think about windows. Internet explored has the ability to download active-x programs and execute them automatically. The programs are actual binary executable programs that run intel code. Now,on top of that,take internet explorer,and hook it so deep into the os that its hard to tell where the os stops and the application begins. Now your active x programs can not only mess with the app,they can mess with the os. On top of that,integrate the web browser into the email client so that it displays html code in a preview pane automatically when it receives an email.n Each of those problems can be secured,but as soon as one fix is made,someone else has found another way to exploit it.There is no way to fix it because there are just too many ways to exploit it.When you think about it,linux and osx both have these same issues lurking. They do the same things,its just there isnt enough market for them to be worthwhile to hack. The real problem is not that windows has a particular buffer overrun issue or a way to trick windows into running something as admin,its that features were added which are very convenient,but no one questioned whether they were going to make a giant security hole. Since apple and the linux developer community have been copying the windows model ,so that they can have similar functionality to windows,they too have inherited the problem. (yes,linux and osx copy windows. Its of course true that windows copies them as well,but thats neither here nor there. By copying the windows features,we inherit some major security flaws.) Unix of course has its share of flaws,and has had many more in the past. Hacking was born on unix. In fact,there are programs for unix that let you run programs remotely without any authentication. On many older unix operating systems,these were automatically enabled by default. Programmers need to wise up and before adding or changing any features,the first question they should ask themselves is,"how could I use this to cause some damage"
|
Mashie Saldana
Hooligans Of War
|
Posted - 2007.09.04 09:30:00 -
[11]
A distributed system like this will not work at all. You need as little latency as possible within the cluster.
Read the bottom part of this devblog, infiniband and RDMA is the only way forward tbh.
Light Assault Launchers & Defender FoF ideas |
| |
|
| Pages: [1] :: one page |
| First page | Previous page | Next page | Last page |