Dear GIBA,
we understand that there are different preferences amongst users and we'll unfortunately not be able to make everyone happy.
Just to give you our perspective: Quantum chemical calcs are usually conducted on high-performance clusters and often run for days on this much more powerful hardware (compared to the average PC). This may give you an impression on the challenge our team faces in designing calcs that are both meaningful and at the same time doable on the WCG with the given time and hardware constraints. We feel that the scientific relevance of this project would be severely limited if we were to go below 12h runtime.
But we appreciate your comment!
Best wishes from your

Harvard CEP team

Dear Jack007,
most of your suggestions are actually no problem in terms of the wus or Q-Chem. But customizations in DC are very hard! It just doesn't lend itself to squeezing out the last bit of performance for a million different computers. I think I said this before - simplicity is a quality in itself!

But thanks for your comments anyways. There are certainly still a number of doable improvements and we are always keen on learning what the users have in mind on this front.

Best wishes from the
Harvard CEP team

from what i understand they normally get one computer (at least) out of the 2 completing all the way through and the partial is used to verify the results.

if both end early they will still have enough data to determine if they want to do more research on that WU.. in that case they run it in house instead of creating a specific special WU and sending it to us.

as for chopping the WU's to 2 blocks instead of one... I do not agree again the high end computers normally get the work done within 8 hours and the slower ones end at 12 hours.

not to mention if they did multiple stages (like DDDT2) they would probably run into the same problem lots of dry spells while they wait to be ready to release the next set of results.

as it is right now they can run lots of results and if there is a gap in there data they can run that gap in house

with dddt2 they end up havering to wait till a large set of work is done and then they can make the next sets.

Personally i like how this project works and i am glad to know that even if one of my machines doesn't complete a job that the results are very useful and there wont be a data gap cause someone else probably finished it

...
Another reason not to do it in multiple jobs is size.
For instance if the results from phase 7 needed for phase 8 are larger than the result needed to be sent back there is a lot larger data files. its kinda like if you have a math test they give you the question and you need to give them the answer you get (may be a single # or a small formula) so the result they need is small.. if they do a stop send back make next WU and put it out they have an question and lots of work steps listed to be sent to there computer as well as sent out again to the next

Huh?!?!? I just got credit for 21 hours CPU time on a CEP2 WU. ????

E201034_ 876_ A.29.C21H12N6S2.284.2.set1d06_ 2-- kate-jetway64 Valid 1/28/11 13:10:22 1/29/11 16:58:54 21.03 102.3 / 63.8

This WU had problems -- just as job #2 was about to finish, BOINC tried to contact the Internet when my connection was down, which bounced it back to the start of job #2. Then something else happened much later. Here's the end of the log.

[09:13:29] Qink name = drvman
[09:19:52] Qink name = optman
[09:19:52] Qink name = anlman
[09:24:46] End of Job
[09:24:47] Finished Job #2
[09:24:47] Starting job 3,CPU time has been restored to 40500.690000.
Killing job because cpu time limit has been exceeded. 40500.690000||35220.580000||0.000000
[ERROR] Failed to open either source or destination files while copying A.29.C21H12N6S2.284.2.bp86.svp.n.bp86.svp.n.sp/stdout.txt to A.29.C21H12N6S2.284.2.bp86.svp.n.bp86.svp.n.sp.out. Error: 2
[09:24:47] Finished Job #3
09:24:50 (14171): called boinc_finish

</stderr_txt>
]]>

This is a very slow machine (D525 Atom). I don't have it run much CEP2, but when it does, it's always stopped at 12 hrs, usually around job #8. I think thus WU could have been in progress for 21 hrs. Obviously I was not watching it -- I knew about the Internet connection interruption because I had to get the connection re-established when I came home.

dkt, kateiacy's looks like a valid job (except for it took 21 hrs). We only see up to job #3 because the time limit finally expired. Also job 14 & 15 are useful, they are not to validate other results, they are more information.

here is an analogy to help you understand. Say you are looking for a date.

the 1st 3 jobs (jobs 0-2) gives you the most important information about a the candidate
Job 0 tells you the person's gender. (easy to get)
Job 1 tells you the person's approximate age. (easy to get)
Job 2 tells you if the person's single (takes more effort, need to actually talk to the person, which sometimes is a LOT more effort)

the next 9 jobs (job 3-11) gives you some more useful information, and are easy to get once you did job 0-2
Job 3 tells you their name
Job 4 tells you their hobby
Job 5 tells you their job
Job 6 tells you their favorite movie
Job 7 tells you their favorite food
Job 8 tells you they like to do in their free time
Job 9 tells you their favorite color
Job 10 tells you their likes
Job 11 tells you their dislikes

the next 4 jobs (job 12-15) give you more detailed information, but are much harder to get
Job 12 tells you their phone number
Job 13 tells you their address
Job 14 tells you their birth date
Job 15 tells you their ring size

so you see, all of the jobs reveal important information, but if you are looking to date a single female around the age of 20, and the candidate (WU) you are looking at is male and 44, then you don't need his phone number.

CEP2 send out all of these WU looking to match a certain profile. The first 3 jobs tell them what if there is a match or not. If a WU is a match for the profile and not all jobs are finished, then the harvard team will re-crunch them in-house


edit : I am wondering if it took 21 hours because job 2 took so long. If you look closely, once it got to job 3, it quit right away. So it might be you have to do jobs 0-2 before you can quit (even if it takes longer then 12 hrs)