Welcome to MilkyWay@home

bypassing server set cache limits

Message boards : News : bypassing server set cache limits
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
NinTrance

Send message
Joined: 21 Aug 10
Posts: 1
Credit: 7,154,675
RAC: 0
Message 46431 - Posted: 4 Mar 2011, 4:05:31 UTC
Last modified: 4 Mar 2011, 4:15:32 UTC

Why not ask if there is some people want to donate hardware?
raid controller? bunch of disks and so forth?
I wouldnt be surprised if someone gladly donated =)
or whatever hardware you need to fix botlenecks
ID: 46431 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ExtraTerrestrial Apes
Avatar

Send message
Joined: 1 Sep 08
Posts: 204
Credit: 219,354,537
RAC: 0
Message 46436 - Posted: 4 Mar 2011, 8:47:41 UTC

I'm not sure this is something which can be solved by faster hardware. With new ATIs crunching WUs almost within a minute and typical BOINC cache settings of at least a few hours any server hardware would get overwhelmed.

MrS
Scanning for our furry friends since Jan 2002
ID: 46436 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
unnefer

Send message
Joined: 5 Sep 10
Posts: 2
Credit: 17,543,582
RAC: 0
Message 46842 - Posted: 30 Mar 2011, 14:45:40 UTC - in response to Message 45739.  
Last modified: 30 Mar 2011, 14:47:15 UTC


If people have large caches, that means the work they're crunching can come from very old versions of those populations which have since evolved quite a bit away from where they were when the user filled up their cache. When they return the results there's a lower chance for the results to improve the population of results we're currently working with.
<br>

So that's why our cache is so low

--Travis

Travis,

This is not exactly true and I'm sure you know it. Ppl using faster GPUs will crunch more workunits in the same time as a person with a slower GPU.

For example, a person with a 5870 can crunch a wu twice as quick as a person with a 4870, yet they get the same amount of maximum taks.

Going by your comment, this in itself would also be a bad thing, because the 5870 will finish all 6 tasks before the 4870 has finished 3 of them, it will then request another 6 tasks and complete those before the 4870 has finished its fist 6 tasks. So the 5870 is going to crunching "newer" versions of those populations and the 4870 will be crunching "older" versions of the populations and in effect they will be contributing to the exact thing you do not want to ocurr.

You (mw@h) can penalise the faster gpus as much as you like, but regardless of the wu limit, at any given point in time the faster gpus will be on the "newer" populations while the slower card is still crunching "older" populations.


Rather then penalise the faster gpus, you (mw@h) needs to penalise the slower gpus, by either reducing the maximum tasks allowed for the slower gpus or increasing the maximum tasks allowed for the faster gpus. That way, both the slower and the faster cards are always crunching the same "aged" populations.

Ideally, you (mw@h) should determine the maximum amount of tasks based on the actual gpu series being used, and then adjust the maximum tasks allowed based on the relative performance between the gpu series.

Lastly, I understand that your server may not be able to validate more then 500K workunits per day, but you (mw@h) also need to understand that it is actually costs us us members quite a bit of money (in electricity) to crunch those workunits for you (mw@h).

And those of us who actually need a higher task limit do not appreciate having our rigs sit their doing nothing because our BOINC rigs have run out of workunits to process - to have them sit their doing nothing still costing us money.

Food for thought.
ID: 46842 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Ed.T

Send message
Joined: 1 Feb 11
Posts: 17
Credit: 16,245,184
RAC: 0
Message 46843 - Posted: 30 Mar 2011, 17:11:12 UTC - in response to Message 46842.  

Really!? Once it gets to the point where there's no work to fetch then what's the difference between my system being out of work and your system being out of work?

If fact, I bet your (newer) system idles more efficiently than mine so from an energy-use-on-this-planet point of view it seems more efficient for your system to sit idle than mine.

Just thinking...

- Ed.
Please: WCG - Help Cure Muscular Dystrophy
ID: 46843 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile banditwolf
Avatar

Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0
Message 46856 - Posted: 31 Mar 2011, 3:02:04 UTC - in response to Message 46431.  

Why not ask if there is some people want to donate hardware?
raid controller? bunch of disks and so forth?
I wouldnt be surprised if someone gladly donated =)
or whatever hardware you need to fix botlenecks


It has been suggested and offered numerous times from different people over the years.
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.
ID: 46856 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2

Message boards : News : bypassing server set cache limits

©2024 Astroinformatics Group