Upcoming update

[Stardragon]

There is a client/screensaver update scheduled for Thursday, December 18, 2003, at 10:00 am EST. Now that the algorithm changes have been tested, we would like to release the update with sufficient time to ensure a smooth transition before the holidays, despite the fact that we have not yet reached the posted 5 billion target.

[Paratima]

Thank you, thank you, thank you!

We appreciates it, we does.

[dbuske]

What does the update include? Is it generations of 100 and ability to MAYBE get more similar proteins to the native one?

[Paratima]

Yep.

[pfb]

So will this mean the server will allocate points based on the 100 strucs/gen? Or will we still be using the current 50 strucs/gen despite processing 100 strucs/gen?

Will the target until the next protein changover be amended? Or are we still aiming to hit 5,000,000,000 somtime in early January 04?

[Rebels Haven]

Will this new client auto-update?

[Brian the Fist]

Yes it will autoupdate, and yes you will get scored for the full 100 strucs per generation, not 50 (assuming we remember to update the server accordingly..)

[Mikus]

Originally posted by Brian the Fist
Yes it will autoupdate, and yes you will get scored for the full 100 strucs per generation, not 50 (assuming we remember to update the server accordingly..)

I think that the points system should reflect the effort, instead of the amount of structures traversed.

The first beta, though doing 100 structures per generation, in my experience took LESS elapsed time (i.e., CPU cycles) than the normal client doing 50 structures per generation. Meaning that somebody running the first beta appeared to receive MORE points per day than somebody running the normal client. Now, if the points for that first beta had been multiplied by 100 rather than by 50, the somebody running the first beta would have accumulated points twice as fast as the somebody running the normal client. I do not think that would have been fair.

The second beta seems to be slow. I do not know if it is twice as slow as the normal client. But if it is NOT twice as slow, then in my opinion multiplying by 100 will result in current participants accumulating points faster than previous participants, who earned their points on the basis of multiplying by 50.

Do you consider it o.k. if the "average points awarded per day" for the new client are higher than for the client previously used ?

mikus

[MarcyDarcy]

Are we, with this new client version, going to chew on the same protein?
2nd question. Is the timelimit rule (24 hrs, 48 hrs) set for this clientupdate...must everyone update their client and will results of the old client being denied and givin no credit for?

/questions over and out

[deranged128[OCAU]]

Another question would be, "Will the points awarded for work done with Beta clients also be scored under the 100* rule if they are uploaded after the update?" (Friday morning here).

I am also receiving an error when my local update server tries to log onto the offical one.

Server.status file is showing:

130 116 118 5000 130 116 118
SERVER DOWN FOR MAINTENANCE, TRY AGAIN LATER
ftp.mshri.on.ca /pub/distribfold/download/patch/
www.teamstirfry.net /distribupdate/

File date stamp is 17/12/2003 09:24

and this from the error.log:

ERROR: [000.000] {foldtrajupdateproxy.c, line 125} Unable to check update file length
ERROR: [000.000] {foldtrajupdateproxy.c, line 310} Unable to get size of update file /distribupdate/distribfold-patch-108-FreeBSD.tar.gz from www.teamstirfry.net
ERROR: [000.000] {ncbi_socket.c, line 1258} [SOCK::s_Connect] Failed pending connect to ftp.mshri.on.ca:80 (Unknown) {errno=Invalid argument}
ERROR: [000.000] {ncbi_connutil.c, line 801} [URL_Connect] Socket connect to ftp.mshri.on.ca:80 failed: Unknown
ERROR: [000.000] {ncbi_socket.c, line 1258} [SOCK::s_Connect] Failed pending connect to ftp.mshri.on.ca:80 (Unknown) {errno=Invalid argument}
ERROR: [000.000] {ncbi_connutil.c, line 801} [URL_Connect] Socket connect to ftp.mshri.on.ca:80 failed: Unknown
ERROR: [000.000] {ncbi_socket.c, line 1258} [SOCK::s_Connect] Failed pending connect to www.distributedfolding.org:80 (Unknown) {errno=Invalid argument}
ERROR: [000.000] {ncbi_connutil.c, line 801} [URL_Connect] Socket connect to www.distributedfolding.org:80 failed: Unknown
ERROR: [000.000] {ncbi_socket.c, line 1258} [SOCK::s_Connect] Failed pending connect to www.distributedfolding.org:80 (Unknown) {errno=Invalid argument}
ERROR: [000.000] {ncbi_connutil.c, line 801} [URL_Connect] Socket connect to www.distributedfolding.org:80 failed: Unknown
ERROR: [000.000] {ncbi_socket.c, line 1258} [SOCK::s_Connect] Failed pending connect to www.distributedfolding.org:80 (Unknown) {errno=Invalid argument}
ERROR: [000.000] {ncbi_connutil.c, line 801} [URL_Connect] Socket connect to www.distributedfolding.org:80 failed: Unknown
ERROR: [000.000] {ncbi_http_connector.c, line 101} [HTTP] Too many failed attempts, giving up
ERROR: [001.001] {foldtrajupdateproxy.c, line 665} Unable to get latest version info, please check network connection

How long will the server be down for maintnenance or should I turn off the local update? Is the problem my end? The PC it's running on has no trouble acting as my mail server and runs the DF client locally too. Internet is 1.5 Mbit ADSL

[Anteraan]

Mikus, you raise some good points. Allow me to throw out my $.02, just to add to the discussion.

Originally posted by Mikus
I think that the points system should reflect the effort, instead of the amount of structures traversed.

While I agree completely in principle, we're far past the point where that can happen, imo. When I first started folding, proteins tended to be smaller and went much more quickly than the ones I've seen recently. Clients are going to change, and compensating for that might be more work than it's worth. I'm not sure there can ever be true "fairness" among the scoring, unless we start counting clock cycles (and even then there would be AMD vs Intel issues).

The first beta, though doing 100 structures per generation, in my experience took LESS elapsed time (i.e., CPU cycles) than the normal client doing 50 structures per generation.

I'll agree with this, the first beta was blazing. I loved it, and it was amazing to watch. Although incredibly efficient, I guess it didn't provide the information desired, however.

Meaning that somebody running the first beta appeared to receive MORE points per day than somebody running the normal client. Now, if the points for that first beta had been multiplied by 100 rather than by 50, the somebody running the first beta would have accumulated points twice as fast as the somebody running the normal client. I do not think that would have been fair.

I can see your point, completely. However, I'm not averse to some form of statistical reward for those who choose to help beta-test a new client. Double points for a few days, in the grand scheme of things, doesn't seem to out of line. If there was a new client every few weeks, that would be a different story. Also keep in mind that since the first beta was not scored at 100*(sqrt gen), it's not a huge issue anyway. :P

The second beta seems to be slow. I do not know if it is twice as slow as the normal client. But if it is NOT twice as slow, then in my opinion multiplying by 100 will result in current participants accumulating points faster than previous participants, who earned their points on the basis of multiplying by 50.

It sure seems slow, and it might just be twice as slow. I'll leave that for others to determine. Even if it's not quite twice as slow, I don't have a major issue with the 100* factor. Like I mentioned above, proteins and clients change. The only people who are going to get "caught up to" are the ones who have left the project. If they've left the project they either:
a) don't care about the project/stats anymore or,
b) will keep an eye on the situation and come back when they see statistical improvement.
Either way, the opportunity is equal in the present situation, and the overall situation (in terms of the project) is certainly dynamic.

Do you consider it o.k. if the "average points awarded per day" for the new client are higher than for the client previously used ?

Good question, and a fair question. I'd be curious to the answer given as well. I think both answers have strong arguments too. Consistency is the obvious stance for someone who replies "no" to your question. Then again, it would certainly not be unprecedented for the DF scoring system to reward people for sticking with it. Heck, the scoring system is biased so that getting to higher generation numbers is rewarded. One could use that argument (as a microcosm of the DF project) to justify the 100* factor, even if it meant an increase in average points awarded per day. That said, I don't think it would be right to have a HUGE increase in points per day either. Keep in mind that this measure also assumes no change in the overall computing power in DF. Hopefully, with Christmas coming, many folders are going to add to/upgrade their farms, which could be quite a confounding factor. "Santa, where is my dually???"

In the end, which answer is right? I don't know, I just run the program to help out.

[pfb]

In reply to those saying Test client #2 is slow - I have been running it on 4 PCs now (1 Linux, 1 Win2000 and 1 WinXP Pro (Dual processor)) and have seen it run at the same, if not slightly faster, than the first test client...without the plateauing affect (although it's still there but only for a couple of gens, not 100+)...

[deranged128[OCAU]]

Originally posted by pfb
In reply to those saying Test client #2 is slow - I have been running it on 4 PCs now (1 Linux, 1 Win2000 and 1 WinXP Pro (Dual processor)) and have seen it run at the same, if not slightly faster, than the first test client...without the plateauing affect (although it's still there but only for a couple of gens, not 100+)...

I've been running the two test clients on identical P4 2.4b's for the past 5 days and have found there to be little difference in speed. The changes implemented in the 2nd Beta were for how it calculates energy, not how it folds.

We all know the protein produces some very variable gen times so even 5 days is probably not enough to really test it.

[PCZ]

ALL
I find the second beta to be just as fast as the first.

[pointwood]

I believe it is quite okay if those running the beta gets more points, they are after all running a risk by running a beta and no one was excluded from downloading and running the beta.

OTOH, I didn't run the beta client to get more points, I installed it to help the project as much as I can.

It is difficult (if not impossible) to make stats that are 100% fair, but I think the DF stats are pretty okay in that regard. It would be quite difficult to make a system that's more fair than the current system.

[Stardragon]

The protein will stay the same in order for us to evaluate the algorithm changes. The 24/48 hour rule still applies for old results, as usual.

[pfb]

Originally posted by Stardragon
The protein will stay the same in order for us to evaluate the algorithm changes. The 24/48 hour rule still applies for old results, as usual.

So will the current protein scores be reset in order to clearly see whether the changes have been worthwhile? Or are the current top 10 RMSs staying as they are? I would have they reset as we are doing a protein update (just happens to be the same protein)...but just wanted to clarify

[Brian the Fist]

As far as you, the users, are concerned, this will be like any other normal protein update (with the executable being updated as well). We hope this new energy function folds the protein better but we won't know until we try it on a large scale with all of your help. If it performs poorly, we'll try other changes - we still have numerous ideas in the works and in the development stages.

[Grumpy]

I always was a sucker for dangerous assignments

[dbuske]

I am glad the program update is coming out tommorow. I am stuck in some really slow generations. The program is keeping the same shape and trying the same thing over and over again.
Maybe the new update will kickstart the folding again.

[dbuske]

Well it's past 10am and no program update?

[Rebels Haven]

Last time we did an update at 10:00, it wasn't til about 1:00pm that the update started and was available for download.

First clue is one can't get to the download page....

[magnav0x]

MAIN SERVER IS DOWN FOR PROTEIN CHANGEOVER

Please try again later. Press Enter

[Brian the Fist]

Do not fear, the 10 AM is just approximate.. its coming soon

[Rebels Haven]

Would it be possiable to also post the "distribfold-update.exe" file, so that us folks that have no choice then to be on dial-up, don't have to pull out hair out with each and every up-date....?


Even a link in this forum would help...

[pfb]

according to the news page:

The update is delayed due to technical difficulties. It will now take place between 1:00-3:00 pm EST this afternoon (Thursday).

[Rebels Haven]

:sleepy: :sleepy: :sleepy:

[Grumpy]

So will the points be doubled now that it is 100 Structures, if so, when will that come into effect.

[Anteraan]

It appears to have already, Grumpy. I uploaded gen 0, 1, 2, and 3 to check it and found myself with 514 more points.

And yes, that means we're getting 100 for gen. 0 as well.

[Grumpy]

Well, does that mean those who dumped with gens from the old client will get double points ? Because anyone who no netted the last week and now dumped will get a HUGE bonus. If I had have been offline Folding the last week, it would mean and extra 2 Million points for me by dumping now. I hope this is not the case because it will lead to an ugly scene with those who Fold online

Never a quiet moment, whenever a points change is made someone is unhappy, but this will get ugly if true

From the Stats, it does not appear so, but a confirmation from Professor Howard would be nice

[Anteraan]

No need to worry, Grumpy. I was forced to nonet status for a day or so, and no bonus here. Mine were scored with the 50* factor.

[Grumpy]

But did you Upload as it updated, or dump them while the Update was yet to appear..



Well, all is quiet, no one is game to say if their last weeks no net dumps were doubled. If this is the case, us Online Folders will expect this oversight to be fixed like the last few Stats Points Issues were by retrospective correction. Thank you one and all.

[Anteraan]

Originally posted by Grumpy
But did you Upload as it updated, or dump them while the Update was yet to appear..

I actually dumped that particular client with my auto-update. Previous to that, I had downloaded a new (post-update) client, I ran a few gens, waited for the scoring, saw that it was 100* factor, then accessed the older client and put it online. It then dumped and auto-updated. Those points were scored with the 50* factor.

So, in effect, I made certain that the 100* scoring was in effect before I dumped them, yes. If I would have had a huge amount of work stored, it would have been a bonus, but I only kept a few gens for that purpose. I'll never be a real driving force in the stats thing, since I'm running a whopping one box with my XP 2000 CPU (Fear my clock cycles!!), so I do stuff like this to experiment.

[Grumpy]

Was the older Client you dumped with the Test or Regular Client.

[Anteraan]

I dumped a couple with test and a couple with regular. At one point I had 5 diff. client directories on my system, messing with this stuff. In the process of cleaning up my system, I found about 15 gens of old (50 structure) client, so I dumped those too. What I referred to in my previous post were beta gens though, just to clarify.

[Grumpy]

One of our Folders at OCW says he uploaded 600,000 points and got credited with 1.2 M

[Anteraan]

Well, that's interesting. All I'll say is that I had no such luck. If I had gotten double credit, I'd certainly post it, since 2 x squat is still pretty much squat.

I won't say that it didn't (or couldn't have) happened for anyone else. All I can say is that it didn't happen for me. I guess that's the scientist in me. In any case, I hope it all gets resolved. After all, I have stated previously that I'm not opposed to people getting the 100* factor for running the beta client (and that includes the first beta as well), but everyone should get it then. Having to run nonet to get it seems more than a tad sketchy, imo.

[Paratima]

Hey, it's just points!