Well, I think this problem is not as bad atm as it used to be, as the WUs are simply so much larger. My desktop PC used to do a "small" S5R1 result in 45 minutes, which means if it had run all day, with 100% res share, and me not doing something CPU intensive, there would have been 32 results to report. I have no problem imagining that this would have killed the server! But the current WUs take 28 hours on this machine, which means less than 1 result/day... this is in another league...
Of course this isn't foolproof, and someone with a really low resource share will still take a long time to crunch that first WU and get the estimate corrected. But in that case, there won't be a large cache to worry about (in terms of numbers of WUs).
Perhaps. I wasn't so much concerned about a really low resource share as I was a new host or an old host that hasn't connected in a while that had either a 100% share that would grab too much based on the benchmark (new host) or something like 30-50% with a 10-day cache for an old host...
Also, as far as "foolproof", there are hacks that can be done to the xml files as well as simply suspending projects to get more work, but I would guess those would have to be considered "abnormal / bad user behavior"...
Reporting 1 result takes almost as much overhead as reporting multiple results.
There is one thing i have bin wondering about. Is there a difference in the overhead for the database server if ONE computer reports 10 results of its own or 10 results from 10 different hosts (One from each host)?
Then you're really interested in a subject, there is no way to avoid it. You have to read the Manual.
Remind me what conversation we had? I haven't been at Primegrid in eons.
5.5.0 (return_results_immediately) vs. anything else that doesn't (5.6.5 was in use at the time)...and the idea of batching uploads vs. batching reports and disagreeing over the database expense of reporting immediately vs. having a lot of clients out there repeatedly attempting to report...
FYI, I'd still apreciate some kind of response from the project, even if it is in another thread (preferably sticky) about the current status of the main builds (Windows, Linux, Mac)... Something that gives an idea of the amount of remaining issues and when/if optimization may occur. Specifically, and somewhat selfishly I suppose, I'd like to know about the AMD/Windows penalty and what is being done...
Well, I think this problem is
)
Well, I think this problem is not as bad atm as it used to be, as the WUs are simply so much larger. My desktop PC used to do a "small" S5R1 result in 45 minutes, which means if it had run all day, with 100% res share, and me not doing something CPU intensive, there would have been 32 results to report. I have no problem imagining that this would have killed the server! But the current WUs take 28 hours on this machine, which means less than 1 result/day... this is in another league...
RE: Of course this isn't
)
Perhaps. I wasn't so much concerned about a really low resource share as I was a new host or an old host that hasn't connected in a while that had either a 100% share that would grab too much based on the benchmark (new host) or something like 30-50% with a 10-day cache for an old host...
Also, as far as "foolproof", there are hacks that can be done to the xml files as well as simply suspending projects to get more work, but I would guess those would have to be considered "abnormal / bad user behavior"...
RE: When I find something
)
BOINC 6.0, coming in the fall, will be it then, as if you keep running old BOINC you eventually won't get work anymore to immediately report. ;-)
@Ageless, @Brian Ah, I
)
@Ageless, @Brian
Ah, I see, thanks for the explanation!
CU
BRM
RE: RE: When I find
)
What ever became of the discussion with you and miw at Primegrid and did Rom do any more investigation into it?
RE: What ever became of the
)
Remind me what conversation we had? I haven't been at Primegrid in eons.
RE: Reporting 1 result
)
There is one thing i have bin wondering about. Is there a difference in the overhead for the database server if ONE computer reports 10 results of its own or 10 results from 10 different hosts (One from each host)?
Then you're really interested in a subject, there is no way to avoid it. You have to read the Manual.
RE: Remind me what
)
5.5.0 (return_results_immediately) vs. anything else that doesn't (5.6.5 was in use at the time)...and the idea of batching uploads vs. batching reports and disagreeing over the database expense of reporting immediately vs. having a lot of clients out there repeatedly attempting to report...
Never mind Brian, found it
)
Never mind Brian, found it already.
Ziran, take a look at:
RomWorld Discussion The evils of 'Returning Results Immediately'
and
RomWorld update on this
and
The Primegrid thread in question.
FYI, I'd still apreciate some
)
FYI, I'd still apreciate some kind of response from the project, even if it is in another thread (preferably sticky) about the current status of the main builds (Windows, Linux, Mac)... Something that gives an idea of the amount of remaining issues and when/if optimization may occur. Specifically, and somewhat selfishly I suppose, I'd like to know about the AMD/Windows penalty and what is being done...
Thanks...