For those not in the know,
TechIMO is a technical community site closely associated with ResellerRatings. It is a great place to discuss current events in technology, plus it is home to one of the largest organized distributed computing teams on the Internet. The latest version of this information can always be found
here. Anyway, we are always seeking new members for our Find-a-Drug effort, so please take a couple of minutes to read this important post.
Find-a-Drug is a not-for-profit distributed computing project maintained by Treweren Consultants, the developer of the THINK software platform for the basic research or “virtual screening” of potential drug candidates. The objective of the virtual screening process is to ascertain the scientific feasibility of the candidates and their potential value to society. All revenues, including royalties from the sales of any resulting drug development, in excess of costs are donated to research.
Find-a-Drug conducts a series of “project queries” targeted at drugs for diseases which have a major impact on health. Project queries have included or will include potential drug candidates against Cancer, HIV, Pesticides and Herbicides, Proteome, Bioterrorism, Respiratory Diseases, Multiple Sclerosis, Malaria and CJD. Members may freely elect to opt in or out of specific project query families at any time.
The
Find-A-Drug website provides detailed background information on each project query and its progress, plus presents the contributions of all participants in the form of
individual statistics and rankings for
organized teams which many members choose to start or join. Many of the teams consist of members from special-interest forums such as TechIMO, and there are several teams of a national character.
Note, the official Find-a-Drug “stats” are updated four times daily at 06:00, 12:00, 18:00, and 24:00 UTC. There are also other sources for following FaD stats, such as the
easy overview site offered by Wizzard~of~Ozz or the
comprehensive analysis site offered by TA GeoffS.
How to Join the TechIMO FaD Team 1. Download and start the installation of the software.
2. In the setup control panel, under the member details tab, enter the nickname you wish to use - this nickname is what will appear in the Stats – and your e-mail address. You can run multiple installations on different machines. If you use exactly the same nickname and e-mail address, the accounts and results will be merged into one within the stats.
3. Under the projects tab, select the various projects in which you wish to participate.
4. Under the general tab, fill in your country (optional, but desirable to get the fastest server connection), your Internet connection plan (”When Online” strongly recommended), and 2037 under Team Number (compulsory!).
5. Save and quit the setup control panel! After two or three minutes, the THINK software should be up-and-running. You can check this by hovering the mouse pointer over the FaD icon in the system tray. The software will now look after all necessary uploading and downloading automatically - a complete “hands-off” experience if you want it that way.
If you need more detailed help on how to setup and run the software, there is more detailed description in the
Member to Member Support Forum within the FaD boards. Our fellow crunchers, but nevertheless competitors, with Team Mauisun have prepared an excellent illustrated
installation and configuration guide for both the Window and Linux platforms.
How THINK Works
The FaD software THINK (To Have INformation and Knowledge) downloads one or more jobs at a time, depending on your processor speed. Each job consists of a protein target and a set of 100 molecules. When your machine has spare processor cycles, the software generates about 100 derivatives for each molecule, giving a total of 10,100 “sub-molecules.” It then generates all the possible shapes or conformations of each sub-molecule and determines whether it can interact favorably with the receptor in the protein target. A sub-molecule which the software predicts will interact or bind well with the protein receptor is referred to as a “hit.”
The process is illustrated graphically by a ball-and-stick model - right click on the FaD icon in the system tray and select Graphics on the pop-up menu. Each ball corresponds to an atom, and each stick to a chemical bond between two atoms and these are colored in accordance with the elements. The common colors used are: blue for nitrogen, red for oxygen, yellow for sulphur and cyan for carbon. Note, the graphics view options do not work if Open GL Graphics is not selected under the options tab in the setup control panel.
Checking on Job Progress
By hovering your mouse over the FaD icon in the system tray you can track the progress of each job. Progress is expressed as a percentage of the number of sub-molecules completed. At the beginning of a job this is often indistinguishable from the time. However, the more flexible molecules in a job - those with the most conformations - are processed towards the end of the job. These molecules often take longer to run. So towards the end of a job, the percentage of molecules completed no longer reflects the time taken as well as it did earlier in the job, and it is easy to imagine that the program has hung.
If you are concerned about a job and suspect it has hung, right-click on the FaD icon in the system tray and select job info on the pop-up menu. If the time (seconds) is ticking over, everything is working properly. While there, you can also see your current ”CPU rating” in the job info box.
There are also snooze and sleep features on the pop-up menu If you select the snooze button, THINK will restart after 15 minutes (default value). If you select the sleep button, THINK will not restart until you resume it.
If for some reason you believe that THINK is slowing down other applications – not a likely occurrence though - you can try using NOTEPAD (Windows) to text edit the "THINK.ENV" configuration file in the Find-a-Drug folder. If, for example, you add the line THINK_PRIORITY=IDLE:IDLE, that will decrease THINK’s priority to an absolute minimum. The only applications which benefit from this sort of edit are Internet server packages, such as those for web pages and email. Other alternative edits are discussed under configuration and setup issues section of the
official Find-a-Drug FAQ.
Calculation of Points and GFlops
GFlops (Giga Floating Point Operations) are a “new” measure of the processing work performed. They have officially superceded the old points system, which was a more abstract measure of performance, but old habits die hard as many people still prefer to talk and compare performance in terms of points. A 1.2 GHz processor will perform roughly 800 Gflops per hour, and the conversion rate normally used is 8 Gflops equal 1 point.
Points are calculated by multiplying the CPU rating by the number of hours a job took. The CPU rating over the previous 30 minutes or so is shown under job info feature on the pop-up menu, together with the percentage progress and the time used to date in seconds. There is absolutely no correlation whatsoever between the number of hits obtained and the points scored.
Find-a-Drug awards Certificates of Recognition to members when they achieve milestones. These can be
downloaded from the official project site.
Running THINK as a Windows Service
Rpertusio provides an easy solution to installing THINK as a service under the Windows operating system.
Quote:
To install Find-A-Drug service:
1. Start menu -> Run...
2. Paste the following line and press [enter]:
sc create ThinkService binPath= "C:\Program Files\Find-a-drug\loader.exe" start= auto error= ignore DisplayName= "Find-a-Drug Think Service"
To Remove Find-A-Drug service:
1. Start menu -> Run...
2. Paste the following line and press [enter]:
sc delete ThinkService |
Using a Central Queue Server
THINK can be run using a central queue server, i.e. one main computer runs the main FaD package and client computers process jobs by connecting to this server over a Local Area Network. This means, for example, that only the main server computer needs Internet access to download/upload jobs.
Tank has prepared an
illustrated tutorial on how to set this up.
Monitoring Your Nodes with FADSPY
FADSPY is a very neat freeware program written by our industrious FaD friends at TSC! Russia (Blackbird). It allows the user to monitor multiple nodes over a network and provides many additonal features for tracking your output, hits, mollies, CPU ratings, etc.
It allows monitoring of standalone machines (you simply give it the network location of the individual nodes' server.dat file), and it will also monitor a queue server.
Once configured, all the nodes are on the main grid and are color-coded as to their status (stalled, stopped or running), and all the current job details for each active node are shown.
The latest version of FADSPY is
0.48 build 10, and there is a
thread at the Find-a-Drug forum to discuss any questions, bugs, features, and so on.
Running THINK on a Machine with No Internet Access
Ultima has this to say about “sneakernetting”:
Quote:
You have to install FaD on the offline computer also. When you register,
select Offline. On the computer with net access, you go to the Queue (Think Server) and select Receive. Download as many jobs as you want and click OK.
Then on Think Server go to File; Export. Select the jobs you want to export and where you want to save them [e.g. on a floppy].
On your offline computer, go to Queue (Think Server), File, Import and select the exported jobs. It should start crunching.
Most important is, DO NOT delete the exported jobs that are in the Queue on the computer with the net access, otherwise you won't be able to upload them to FaD later.
If you need any more info, you know where to find us.
|
Jrsweger offers additional tips:
Quote:
|
You will have to install FaD on both computers, download on the one connected to the internet and set it up to only connect after confirmation. I would guess that if you don't want FaD running on this computer you will want to either make it sleep or exit out of it after you download. On the other, set it up as offline. You will have to download on the internet connected one then export each file separately to a floppy, go to the other computer then import to that one. Unless I'm doing it wrong, I can only export/import one file at a time so it is time consuming. Also when importing, if the queue is empty you will probably receive an error. I have had to go into Task Manager and end the server then restart it to get it to work again. The imported job (that created the error will be there) then you can import more jobs without a problem. After the jobs complete, you will have to export them again and then import them on the Internet computer to upload them. There's probably a lot more involved than what I just said so if you need further help, private message me, and I'll get back to you.
|
Optimizing THINK for HyperThreading
RobRich has this to say about HyperThreading:
Quote:
Intel Pentium 4 HyperThreading processors have two scheduling units, but the backend execution unit remains relatively the same as previous Netburst chips. Considering FaD streams data continuously, it makes little sense to run two instances of FaD on a HyperThreading-capable system. The resulting output will be the same as single instance, or actually slightly less due to scheduler overhead.
Comprehensive testing shows the Pentium 4 platform generally offers better performance with single-threaded applications when HyperThreading is disabled. Again, this is due to scheduler overhead.
If you are using the system in question as a daily-use desktop, then leave HyperThreading enabled for best multitasking performance, though it is recommended to scale FAD back to one instance so only one scheduler is loaded. If you are using the system as a dedicated FaD cruncher, then disable HyperThreading and set FaD to a single instance for maximum performance.
|
I guess this thread just hasn't been updated in a while. Unfortunately fad has closed down. I guess I shouldn't say unfortunately. They finished all the calculations they have far ahead of schedule and thus there is no work left. I guess we should be thankful the project was such a success, but now...no more fad. If you head over to techimo and look in the distributed computing forum you'll find lots of new great projects including the one I recommend...called boinc. It's basically a single platform taht you can run many projects on (so I run 5 seperate projects....they rotate one at a time an hour a piece...it's pretty cool!). Anyways....if you're interested, head over to techimo and check it out 
.
Linear Mode
