Trying to Understand F@H

[RogersDA]

I understand the general idea behind the distributed computing concept. I just want to clear-up some NOOb issues on F@H. If I am wrong or otherwise misstate something please let me know.

1. A work unit (WU) is a specific "data block" for a project. Each project has several WUs.

2. When processing a WU I should see progress from 0% to (eventually) 100%, and those results are uploaded back to Stanford.

3. Points are received for each completed WU.

4. Based on the Projects screen the total points for a given project is analogous to the Credits column.

5. I am running the following, and have copied the first lines from my second WU. What else do I need to do?
Mac OS 10.6.1
Processor Name: Quad-Core Intel Xeon
Processor Speed: 3 GHz
Number Of Processors: 2
Total Number Of Cores: 8
Graphics Model: NVIDIA GeForce 8800 GT

Here is the results of the second WU so far:

Code: Select all

[13:50:37] + Processing work unit
[13:50:37] At least 4 processors must be requested; read 1.
[13:50:37] Core required: FahCore_a2.exe
[13:50:37] Core found.
[13:50:37] Working on queue slot 02 [November 9 13:50:37 UTC]
[13:50:37] + Working ...
[13:50:37] 
[13:50:37] *------------------------------*
[13:50:37] Folding@Home Gromacs SMP Core
[13:50:37] Version 2.11 (Fri Sep 4 09:50:46 PDT 2009)
[13:50:37] 
[13:50:37] Preparing to commence simulation
[13:50:37] - Ensuring status. Please wait.
[13:50:37] Files status OK
[13:50:38] - Expanded 4821862 -> 24057197 (decompressed 498.9 percent)
[13:50:38] Called DecompressByteArray: compressed_data_size=4821862 data_size=24057197, decompressed_data_size=24057197 diff=0
[13:50:38] - Digital signature verified
[13:50:38] 
[13:50:38] Project: 2671 (Run 24, Clone 94, Gen 127)
[13:50:38] 
[13:50:39] Assembly optimizations on if available.
[13:50:39] Entering M.D.
[13:50:48] un 24, Clone 94, Gen 127)
[13:50:48] 
[13:50:48] Entering M.D.
[13:50:57] Completed 0 out of 250000 steps  (0%)
[13:57:05] Completed 2500 out of 250000 steps  (1%)
[14:03:11] Completed 5000 out of 250000 steps  (2%)
[14:09:15] Completed 7500 out of 250000 steps  (3%)
[14:15:20] Completed 10000 out of 250000 steps  (4%)
[14:21:25] Completed 12500 out of 250000 steps  (5%)
[14:27:28] Completed 15000 out of 250000 steps  (6%)
[14:33:31] Completed 17500 out of 250000 steps  (7%)
[14:39:35] Completed 20000 out of 250000 steps  (8%)
[14:45:38] Completed 22500 out of 250000 steps  (9%)

[JimboPalmer]

You seem to be attempting to run the Symmetric Multi Processor core on a Mac with less than 4 cores. perhaps I am reading that wrong.

[Nathan_P]

I understand the general idea behind the distributed computing concept. I just want to clear-up some NOOb issues on F@H. If I am wrong or otherwise misstate something please let me know.

1. A work unit (WU) is a specific "data block" for a project. Each project has several WUs.

2. When processing a WU I should see progress from 0% to (eventually) 100%, and those results are uploaded back to Stanford.

3. Points are received for each completed WU.

4. Based on the Projects screen the total points for a given project is analogous to the Credits column.

5. I am running the following, and have copied the first lines from my second WU. What else do I need to do?
Mac OS 10.6.1
Processor Name: Quad-Core Intel Xeon
Processor Speed: 3 GHz
Number Of Processors: 2
Total Number Of Cores: 8
Graphics Model: NVIDIA GeForce 8800 GT

Here is the results of the second WU so far:

Code: Select all

[13:50:37] + Processing work unit
[13:50:37] At least 4 processors must be requested; read 1.
[13:50:37] Core required: FahCore_a2.exe
[13:50:37] Core found.
[13:50:37] Working on queue slot 02 [November 9 13:50:37 UTC]
[13:50:37] + Working ...
[13:50:37] 
[13:50:37] *------------------------------*
[13:50:37] Folding@Home Gromacs SMP Core
[13:50:37] Version 2.11 (Fri Sep 4 09:50:46 PDT 2009)
[13:50:37] 
[13:50:37] Preparing to commence simulation
[13:50:37] - Ensuring status. Please wait.
[13:50:37] Files status OK
[13:50:38] - Expanded 4821862 -> 24057197 (decompressed 498.9 percent)
[13:50:38] Called DecompressByteArray: compressed_data_size=4821862 data_size=24057197, decompressed_data_size=24057197 diff=0
[13:50:38] - Digital signature verified
[13:50:38] 
[13:50:38] Project: 2671 (Run 24, Clone 94, Gen 127)
[13:50:38] 
[13:50:39] Assembly optimizations on if available.
[13:50:39] Entering M.D.
[13:50:48] un 24, Clone 94, Gen 127)
[13:50:48] 
[13:50:48] Entering M.D.
[13:50:57] Completed 0 out of 250000 steps  (0%)
[13:57:05] Completed 2500 out of 250000 steps  (1%)
[14:03:11] Completed 5000 out of 250000 steps  (2%)
[14:09:15] Completed 7500 out of 250000 steps  (3%)
[14:15:20] Completed 10000 out of 250000 steps  (4%)
[14:21:25] Completed 12500 out of 250000 steps  (5%)
[14:27:28] Completed 15000 out of 250000 steps  (6%)
[14:33:31] Completed 17500 out of 250000 steps  (7%)
[14:39:35] Completed 20000 out of 250000 steps  (8%)
[14:45:38] Completed 22500 out of 250000 steps  (9%)

@RogersDA

#1 - yes but there are 1000's of WU per project
#2 & #3 Yes
#4 ?? Not quite sure what you are getting at but the stats sites will show you your total points to date based on the number of WU's you have submitted and the points they are worth.
#5 I don't own a mac so I can only help a little - are you using the console client or the OSX installer?

@Jimbo
His system specs suggest he should be thinking -bigadv but i'm with you, it reads like the system is only using one core for folding but his frame times look OK to me - i am assuming that a2 core performance on macs is the same as what it is on the windows VM systems

[RogersDA]

Well, a lot of new terminology so bear with me on these questions, too. As to question 4 I was looking at the point summary website and the Credits column (I am not able to post the link in messages here, yet). I did one WU last night and received 1920 points (at least that's what my summary shows). I thought that the entire project was worth 1920 points, which would be divided into a specific number of WUs. Edit:

Edit: Using the Mac installer and not the console.

Also I have a screen grab of the activity monitor of my machine. It appears to show four processes working on the one current WU. Not sure how significant that is.

[Nathan_P]

Well, a lot of new terminology so bear with me on these questions, too. As to question 4 I was looking at the point summary website and the Credits column (I am not able to post the link in messages here, yet). I did one WU last night and received 1920 points (at least that's what my summary shows). I thought that the entire project was worth 1920 points, which would be divided into a specific number of WUs.

Also I have a screen grab of the activity monitor of my machine. It appears to show four processes working on the one current WU. Not sure how significant that is.

Sure, everyone has to start somewhere so please ask away. The points that you refer to is the credit that you receive for each WU that you return to stanford, if you return 10 WU from project 2671 you get 1920 points per WU making a total of 19,200points.

As for points/stats sites there are 3 main ones that people use - in the banner at the top of the page there is a row labelled stats just above the date/time, Offical is the one that stanford maintains, the other 2 are more sophisticated and allow you to see where you are in your team,teams that could overtake your team, etc. When you get a minute have a look and see which one suits you best.
your screen shot shows that you are running at least 4 processes of the WU which means that the system isn't running at peak although its still pretty good. Depending on which intel xeons are in your mac you could be running 8 or 16 processes but i'm not able to tell you how to get the other processes running correctly. You could try here: http://foldingforum.org/viewforum.php?f=6. There are some good configuration guides on there to help you get the best from your machine. Hopefully one day stanford will also release a client for your 8800gt, Windows users are also able to fold on such graphics hardware and can get aboy 5000 points per day from a card the same as yours.

[kiore]

Also I have a screen grab of the activity monitor of my machine. It appears to show four processes working on the one current WU. Not sure how significant that is.

Looks like an SMP (a multi processor client) running to me a2 core and the points sound right, you should be able to run 2 of them or go for the new BIg Advanced as above.
Nice machine
Perhaps you should be in the Apple section rather than General though for some better advice on maximising.
kiore.

[RogersDA]

Thanks a lot for the insight. I started here rather than up there in the Mac forum to start getting some insight.

More questions:

I am running the Mac installer version. I am also running InCrease and FahMon. Is there any benefit to having both InCrease and FahMon?

How does one go about "run 2 of them or go for the new BIg Advanced as above"? Is this something I configure in foldinhg@home in the system preferences or InCrease? Should I ask this in the Mac forums?

On another note what happens to a project after the expiration date has passed without all the WUs being processed? Does this happen a lot? Are the projects re-queued? Just curious...

[P5-133XL]

Pick the monitoring solution that you prefer.

On a single machine, it is not good to run two -bigadv WU's at the same time. First it slows down the science because it is important to get every single WU done as fast as possible so that the next generation (which depends of the previous generation) of WU's can be created. From a points view, the -bigadv gives out bonus points if you return the WU fast and if you do two simultaneously, both will be slow and you'll get less points on each.

If you want to run -bigadv's then I would suggest that you change the -smp option to "-smp 8" so the SMP client/core will use all eight of your processor's cores. That is a better solution, than running two clients, for everyone concerned.

As to what happens when an expiration date passes: There are actually two deadlines for every WU. When the first one expires, the WU you are working on is assigned to someone else even if you are still working on it. If you complete the WU before the second deadline you will still receive credit for it, as well as the other person that got it too (assuming it gets completed in time). When the second deadline expires, you will not receive credit for the WU you've been working on; the client will cancel the processing of the WU; and then go get another one.

[bruce]

Welcome to the FoldingForum, RogersDA.

I see you've gotten some excellent information, an although it wasn't explicitly stated, your item 3 probably should have said ".... within the deadline." but you've obviously figured that out.

3. Points are received for each completed WU.

More questions:

I am running the Mac installer version. I am also running InCrease and FahMon. Is there any benefit to having both InCrease and FahMon?

How does one go about "run 2 of them or go for the new BIg Advanced as above"? Is this something I configure in foldinhg@home in the system preferences or InCrease? Should I ask this in the Mac forums?

I'm not a Mac expert, but I'm pretty confident that InCrease installs FAH in a different location than the latest installer version so you'll likely have trouble if you mix the two methods. Moreover, the instructions for "run 2 of them" are going to be different. I do suggest the Mac forum for this -- and they can correct anything that I may have said that's incorrect or misleading.

On another note what happens to a project after the expiration date has passed without all the WUs being processed? Does this happen a lot? Are the projects re-queued? Just curious...

Each WU has two deadlines: A Preferred Deadline and a Final Deadline. If someone downloads a WU and the results are not uploaded by the Preferred Deadline, the WU is assumed to be lost and a new copy is issued to someone else. This delays the completion of the project, so it's best to avoid it whenever possible. If you return a result after the Preferred Deadline but before the Final Deadline, you'll still get credit, although the work will probably be duplicated (wasted). (In some cases, the deadllines are set to the same value, so this doesn't apply.)

Almost every project consist of a number of trajectories which are run in parallel. Each trajectory is denoted by a pair of numbers (Run,Clone). Each step along a single trajectory (enoted by the Gen number) must be completed sequeuenally, so even if a WU is lost or otherwise not compelted, it's needed before the next Gen can be issued. (See also viewtopic.php?f=16&t=12087&p=118366#p118366)

[Nathan_P]

Thanks a lot for the insight. I started here rather than up there in the Mac forum to start getting some insight.

More questions:

I am running the Mac installer version. I am also running InCrease and FahMon. Is there any benefit to having both InCrease and FahMon?

How does one go about "run 2 of them or go for the new BIg Advanced as above"? Is this something I configure in foldinhg@home in the system preferences or InCrease? Should I ask this in the Mac forums?

On another note what happens to a project after the expiration date has passed without all the WUs being processed? Does this happen a lot? Are the projects re-queued? Just curious...

-Bigadv are the new massive work units available to mac/linux users with 8 or more processor cores. To be able to do them you first need a passskey from stanford http://folding.stanford.edu/English/FAQ-passkey. Oce you have your pass key you will need to enter it into your client and fold 10 normal smp work units before you are eligable for -bigadv. Once you have done this i recommend visiting the mac and -bigadv area's for advice on how to set it all up.

[bruce]

-Bigadv are the new massive work units available to mac/linux users with 8 or more processor cores. To be able to do them you first need a passskey from stanford http://folding.stanford.edu/English/FAQ-passkey. Oce you have your pass key you will need to enter it into your client and fold 10 normal smp work units before you are eligable for -bigadv. Once you have done this i recommend visiting the mac and -bigadv area's for advice on how to set it all up.

Be sure to read and understand the details in the announcement: viewtopic.php?t=10697

[RogersDA]

Thanks to everyone for dealing with my NOOB questions.

One of the first things I did was get a passkey and put that in the appropriate field. Now I known why I need it!

I have since stopped the F@H client in System - Preferences and am going with InCrease. I followed a few steps and now have 8 cores processing



Each step in processing the WU has dropped to about 3.5 to 3.8 minutes.

Now - InCrease in configured with the UserName and Team Number. Will I not be able to go to BigAdv using only InCrease (there is no field in InCrease for providing the passkey)?

[whynot]

3. Points are received for each completed WU.

I just want to highlight one unobvious issue. You (the donor) are for stats system Nick+Team. The credit is given to you (Nick+Team) and your team. You change your Nick then you start from nil; you change your team then you start from nil again. There is no way to take your amassed credit and move it to another team or another nick. Sometime there (in the forum) are requests for such transfers (I think there're PMs and private e-mails too) -- they all are rejected. Not because the people behind F@H's stats are that rude. It's just policy.

[whynot]

Behind the scene I was pointed that I've made a terrible mistake: there's no Nick+Team; points for Nick are points for Nick; points for Team are points for Team (noTeam is still Team). Thus one can change teams at will and still keeping her/his points. However there's no way to transfer points from Nick0 to Nick1 and/or Team0 to Team1. And let's it to be this way.