Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 32
Posts: 32   Pages: 4   [ Previous Page | 1 2 3 4 | Next Page ]
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 4739 times and has 31 replies Next Thread
knreed
Former World Community Grid Tech
Joined: Nov 8, 2004
Post Count: 4504
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Fast work units

Yes - uplinger's 'chopper' has been a very effective test of our scalability and as happened in late December (when we hit our previous record) we are finding some slow spots that we are resolving.

Over the past 24 hours the BOINC grid processed ~356,000 results with an average of 3.7 hours per result (across all projects). I made some adjustments over weekend and making a few more now which should clear up some of these bottlenecks.

Just to give you an idea about the increase in load - the same 24 hour period last week saw ~252,000 results processed through BOINC with an average of 5.3 hours per result.

I have used my 'joystick' to increase the distribution of HCC and HPF2 while these adjustments are made to reduce the volume. Once we get the changes in place we will revert back to the previous settings.
[Feb 18, 2008 3:43:05 PM]   Link   Report threatening or abusive post: please login first  Go to top 
retsof
Former Community Advisor
USA
Joined: Jul 31, 2005
Post Count: 6824
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Fast work units

Please check that joystick again ... no work coming ... from ANY project
----------------------------------------
SUPPORT ADVISOR
Work+GPU i7 8700 12threads
School i7 4770 8threads
Default+GPU Ryzen 7 3700X 16threads
Ryzen 7 3800X 16 threads
Ryzen 9 3900X 24threads
Home i7 3540M 4threads50%
----------------------------------------
[Edit 1 times, last edit by retsof at Feb 19, 2008 2:18:55 PM]
[Feb 19, 2008 2:18:21 PM]   Link   Report threatening or abusive post: please login first  Go to top 
knreed
Former World Community Grid Tech
Joined: Nov 8, 2004
Post Count: 4504
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Fast work units

Work is flowing but there are occasional periods where the 'feeder' is waiting on some queries that are taking awhile to run so for a 15-20 minute stretch of time most clients do not get work.

We are working on the problem.
[Feb 19, 2008 3:08:28 PM]   Link   Report threatening or abusive post: please login first  Go to top 
mgl_ALPerryman
FightAIDS@Home, GO Fight Against Malaria and OpenZika Scientist
USA
Joined: Aug 25, 2007
Post Count: 283
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Fast work units

Hi Everybody,

These quick work units seem to involve the little fragments that I made by splitting other compounds into 3 or 4 small pieces. Since AutoDock uses a genetic algorithm, these little fragments tend to converge on their best docked position and conformation quicker than the bigger compounds.

The work units that I will start creating and submitting for the next several weeks will involve larger compounds, and they should take the usual, longer time to complete.

FYI,
Dr. Alex L Perryman
[Feb 19, 2008 4:44:17 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Papa3
Senior Cruncher
Joined: Apr 23, 2006
Post Count: 360
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Fast work units

So how come the Acetamide units are only 2 hours long? Methinks the chopper is chopping WAY too fine! Let's return to the 12+ hour units please!

Hi Everybody,

These quick work units seem to involve the little fragments that I made by splitting other compounds into 3 or 4 small pieces. Since AutoDock uses a genetic algorithm, these little fragments tend to converge on their best docked position and conformation quicker than the bigger compounds.

The work units that I will start creating and submitting for the next several weeks will involve larger compounds, and they should take the usual, longer time to complete.

FYI,
Dr. Alex L Perryman

[Feb 27, 2008 12:57:51 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Sekerob
Ace Cruncher
Joined: Jul 24, 2005
Post Count: 20043
Status: Offline
Reply to this Post  Reply with Quote 
Re: Fast work units

They should have flushed from the system and normal length workunits should be back in circulation. Fetched some new 'IndazoleNO2', so will report if they are not and still short.

There was a little mishap with the flop estimate of the new HPF2 jobs that have 18 day deadline, causing the estimates to be overall to be way to low after processing 1. These last FAAH show an hour in the Time to Complete column here.

Edit: hmmmm may have to retract the story on FAAH. Yesterday's job average was 2.37 hours. Way too low!
----------------------------------------
WCG Global & Research > Make Proposal Help: Start Here!
Please help to make the Forums an enjoyable experience for All!
----------------------------------------
[Edit 1 times, last edit by Sekerob at Feb 27, 2008 7:45:12 AM]
[Feb 27, 2008 7:41:04 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Sekerob
Ace Cruncher
Joined: Jul 24, 2005
Post Count: 20043
Status: Offline
Reply to this Post  Reply with Quote 
Re: Fast work units

Okay, first IndazoleNO2 is done and took 3.37 hours which is 'normal' on this machine.
----------------------------------------
WCG Global & Research > Make Proposal Help: Start Here!
Please help to make the Forums an enjoyable experience for All!
[Feb 27, 2008 8:51:13 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Papa3
Senior Cruncher
Joined: Apr 23, 2006
Post Count: 360
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Fast work units

That's still very short. Normal Faah work units run 12-14 hours. A 3-hour job length has 4-5 times the network overhead of a 14-hour work unit, since 4-5 work units must now be downloaded to get the same run time as before.
Okay, first IndazoleNO2 is done and took 3.37 hours which is 'normal' on this machine.

[Feb 27, 2008 8:34:51 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Sekerob
Ace Cruncher
Joined: Jul 24, 2005
Post Count: 20043
Status: Offline
Reply to this Post  Reply with Quote 
Re: Fast work units

Hello, but 3.37 hours is on the Ferrari. The same job on my Beetle would take 10. The WCG average goal is 7 hours so if it consistently stays this low, more work will be packed in future batches.

The transmission overhead for FAAH jobs is absolutely minimal. Like 200k per job. That's nothing much these days even on 56k dial up.
----------------------------------------
WCG Global & Research > Make Proposal Help: Start Here!
Please help to make the Forums an enjoyable experience for All!
[Feb 27, 2008 8:44:30 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Papa3
Senior Cruncher
Joined: Apr 23, 2006
Post Count: 360
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Fast work units

OK, let's do a few calculations...

56Kbps = 56K bits per second; 1 byte (character) = 8 bits

The "frame" bits and other communications overhead consumes a significant amount of capacity, so let's assume that 48K bits are actually available.

48K / 8 = 6K per second = 360K per minute.

Let's assume that the average Faah job creates 180K of network overhead. Then each Faah job claims the full capacity of a 56K line for 30 seconds.

If the Faah job runs 12 hours, then that's 1 minute of network overhead per day, about 30 minutes per month.

If the Faah job runs only 2 hours, that's 6 minutes per day, 180 minutes (3 hours) per month.

Users who have to sit there for 6 minutes each day (and several times longer if they only connect once every few days) just staring at the screen while Faah gobbles up their network connection will spend that time entertaining the thought of departing Faah and uninstalling BOINC.

Distributed computing projects are guests on other people's computers. As a guest, it is wise to minimize the extent to which the host will be annoyed by your presence, especially if (as with Faah) you hope to stay forever.
[Mar 2, 2008 12:25:10 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Posts: 32   Pages: 4   [ Previous Page | 1 2 3 4 | Next Page ]
[ Jump to Last Post ]
Post new Thread