Concerns

[alpha]

Over the past couple of days I decided I would do a little investigation as to why the OGR projects hosted by distributed.net are taking so long to complete.

According to the FAQ (http://n0cgi.distributed.net/faq/cache/150.html), OGR-24 consists of ~6,000,000 '5-stubs' (I think that means the first 5 marks are decided by the server, with OGR-25 its 6) ranging from ~900,000 nodes to 150,000,000,000 nodes. From the stats pages there is no way of telling how many nodes have been completed, and/or how many are left.

Then I found some graphs (http://n0cgi.distributed.net/statistics/ogr/). Looking down that page, you can see a graph for OGR-24 percent done (1st and 2nd passes). The X axis is the date, and you can clearly see it is the year 2000. Reading from the graph, in December 2000 the 2nd pass was >70% complete (also note the production from Aug-Dec for 2nd pass. 35% was completed in 4 months)! So what have we been doing all this time?

You can also see an OGR-25 graph that shows in Dec 2000 it was >70% done. That was the 1st pass only, though.

So I'm still wondering how far along we are, and why we are still crunching if we were near completion by the end of 2000. There are relevant questions and answers in the FAQ about why there isn't a completion percentage on the stats pages, but it would be nice to have an update after more than 2 years.

Any thoughts peeps?

[wirthi]

Hi,

alpha, I guess you have missed Bruce Wilsons statement made in Oct. 2002 in the official d.net RC5 mailing list:

Status of OGR
* The OGR client has a bug which weakens the accuracy of our results.
A truncation error causes rulers with many segments to be aborted
incorrectly. We were making some good progress towards fixing this bug
and releasing new clients when we solved RC5-64. At that point, all OGR
development was put on hold in favor of RC5-72.
The full statement can be found in d.net's mailing list archive: http://lists.distributed.net/hypermail/rc5.Oct2002/0061.html

That's the main reason why I don't run OGR right now - it would be my favourite project, if it produced correct results.

Greets,
Wirthi

[alpha]

Hmm, yeah. I remember reading that not so long ago but it didn't occur to me when I started this thread.

So there will be a new client eventually, and it will be used to repeat all the work we have done so far, as a kind of verification. Yes? But October 2002 was a while ago. I would be interested to hear what has been happening since then - like if any progress has been made on producing a fixed client.


That's the main reason why I don't run OGR right now - it would be my favourite project, if it produced correct results.

Just thought I would point out that nobody is saying that we are producing useless (or incorrect) results, only that they will need to be confirmed as correct.

[IronBits]

Well, I'll take mine off the project until more recent information comes forth to clarify everything.
Of course, this won't make Phil very happy over on RC5-72 :p :)

[alpha]

Thats right, you stay back there in 12th position, where you belong :D

[IronBits]

:rotfl:
That damn psecky Phil on RC5-72 :bang:
and Condor on SOB trying to beat me to the 2TcEM spot :swear:
/me mumbles and wonders if he can get the HDTV decoder to crunch :)

[TheOtherPhil]

Originally posted by IronBits
:rotfl:
That damn psecky Phil on RC5-72 :bang:
and Condor on SOB trying to beat me to the 2TcEM spot :swear:
/me mumbles and wonders if he can get the HDTV decoder to crunch :)


LOL...I was just looking at my old Amiga A1200 sat in the cupboard wondering how many k/s I'd get :D

[rus6]

Originally posted by wirthi
Hi,

alpha, I guess you have missed Bruce Wilsons statement made in Oct. 2002 in the official d.net RC5 mailing list:

The full statement can be found in d.net's mailing list archive: http://lists.distributed.net/hypermail/rc5.Oct2002/0061.html

That's the main reason why I don't run OGR right now - it would be my favourite project, if it produced correct results.

Greets,
Wirthi

Some time back I read that they are currently re-doing OGR 25 work, with a new bug fixed client (I'm sure the fix has been in for some time) and compairing the results to that of previous runs, to try and work out how bad the initial runs results actually were.

So I think the current crunching is worthwhile..

[wirthi]

Originally posted by alpha
Just thought I would point out that nobody is saying that we are producing useless (or incorrect) results, only that they will need to be confirmed as correct.
Yes, you are correct; but as you wrote in your first post, all nodes should have been completed by now. So right now the stubs are calulated again and again, and the client is still not corrected.

As soon as there is a corrected client, all results will be double-checked. If both the new (corrected) and the old (buggy) core produce the same result (meaning: the same number of nodes), the result is accepted. If the number differs, then the stub is checked a third time (with the corrected client), so there are two equal results.

Greets,
Wirthi

[alpha]

There seems to be some confusion as to whether we are in the second pass of OGR-25 or not (i.e. whether a fixed client has been released). rus6 posted a link (http://lists.distributed.net/hypermail/rc5.Oct2002/0128.html) in this forum a while ago, which seems to yield some useful information (though outdated). Rather than posting quotes and my opinions, I'd like to see what you guys make of the information in that post. There seems to be some contradiction going on. Damn stupid outdated information :mad:

[PY 222]

So, are we still doing work that is useful or useless?

[rus6]

That'll be my swiss cheese brain again. :D Forgetting what I've already posted, doh!

I still think the crunching is worthwhile.. Pass 1 needs to be completed, regardless of whether the client is buggy or not...

The last I heard from dnet was that they have run into database trouble and couldn't get a large query to run against the results database...

That was a month or so back in the .plans..

Perhaps this is the query that'll check if the first pass is correct!

Hope so...

[alpha]

rus6: Yeah, I read about that. Check here (http://n0cgi.distributed.net/cgi/dnet-finger.cgi?user=nerf). It just seems like a case of waiting.

PY: I would say that we are producing useful work, its just that it will need to be confirmed. So obviously this isn't a very efficient method (it wasn't intended), but it will have to be done (unless someone else takes on the same project and does it in one go :))

To be honest I'd like to see OGR-24 and 25 completed and start something anew, OGR-26 - or something. Things seem to be dragging and going a bit stale. I've developed an interest in OGRs and look forward to what the future holds.

[PY 222]

In that case, I shall move back my main box to OGR25 from RC5!

Crunch on people. We need to move Free-DC to the TOP 100 in the team rankings :help:

[Darkness Productions]

What place are you all in currently?


Originally posted by PY 222
In that case, I shall move back my main box to OGR25 from RC5!

Crunch on people. We need to move Free-DC to the TOP 100 in the team rankings :help:

[the-mk]

Originally posted by Darkness Productions
What place are you all in currently?

Data shown reflects all blocks received as of 17-Mar-2003 at 23:59 UTC

851

climbing up about 5 to 15 places per day (depends on everyone's production...)

OK, who stays at OGR-25 and who moves to another project?

[PY 222]

I never really left because I still have one machine on OGR!

But I have added a dually 1800+ back to the project.

Crunch on people :thumbs:

[alpha]

Originally posted by the-mk
OK, who stays at OGR-25 and who moves to another project?

I'm more or less settled down with OGR now, it seems to have become my 'home' project that I always return to. After recent research and discussion I don't think I'll be crunching for another project for a while. I'm here to stay - we have a goal to reach (as do I :)). Crunch on.

[PY 222]

Yeah Baby... I love the way you think :thumbs:

Lets get the show on the road and push Free-DC to the Top 100!

[IronBits]

I'm not here now, but, I'll be back :)

[the-mk]

Originally posted by IronBits
I'm not here now...

Sorry to read that :cry:


...but, I'll be back :)

Yeah, that's great! You're always welcome! :jester: :cheers:
(all the others are welcome too)

Enough people to decide to stay too :D

I think I should create a "day-counter" when we'll arrive at position 100... better not, because if I see how long it takes I'll get bored or something :( but if some of you want to have a day-counter just say yes!

[wirthi]

Originally posted by alpha
rus6: Yeah, I read about that. Check here (http://n0cgi.distributed.net/cgi/dnet-finger.cgi?user=nerf). It just seems like a case of waiting.

PY: I would say that we are producing useful work, its just that it will need to be confirmed. So obviously this isn't a very efficient method (it wasn't intended), but it will have to be done (unless someone else takes on the same project and does it in one go :))
I don't think so. We ARE doing the second pass of both OGR 24 and 25, but still using the buggy client. This means we will have to check all nodes at least one other time with the corrected client (as soon as it is published).

Since the project is running for a very long time now I estimate we have worked on all nodes at least one time; of course there are some stubs that have not been completed yet (sobody downloaded it, but didn't complete it), but most of them should have been completed. So, in my eyes, checking the stub-space again and again with the buggy client. Everybody running OGR now is just wasting ressources in my eyes (or just trying to improve one's stats).

Don't take me wrong; I really like OGR. I just want to have a realiable client that produces reliable results. Calculating nonsense can't be the aim of distributed.net

[ECL]

My understanding is that the December release (477) fixed the bug, which is the only reason I've done any OGR work since then. I can't find where I read that this was the fixed client, though. I'm pretty sure I wasn't dreaming. Does anybody know someone at dnet who might be able to clear this issue up? We're rapidly running out of valid DC projects.

[alpha]

Earlier in this thread I mentioned a link that rus6 posted dated Oct 2002 which seems to clear up a couple of things.


There is no client available today which is not vulnerable to the truncation bug.

It also states:


Until we get that client out, we are stuck on Pass 1, but Pass 1 is not done yet either.

I don't understand what is taking so long if we are still on the first pass. I agree with ECL - up to date news from someone at dnet would be very nice right now.

[Beyond]

Originally posted by alpha
Over the past couple of days I decided I would do a little investigation as to why the OGR projects hosted by distributed.net are taking so long to complete.

According to the FAQ (http://n0cgi.distributed.net/faq/cache/150.html), OGR-24 consists of ~6,000,000 '5-stubs' (I think that means the first 5 marks are decided by the server, with OGR-25 its 6) ranging from ~900,000 nodes to 150,000,000,000 nodes. From the stats pages there is no way of telling how many nodes have been completed, and/or how many are left.

Then I found some graphs (http://n0cgi.distributed.net/statistics/ogr/). Looking down that page, you can see a graph for OGR-24 percent done (1st and 2nd passes). The X axis is the date, and you can clearly see it is the year 2000. Reading from the graph, in December 2000 the 2nd pass was >70% complete (also note the production from Aug-Dec for 2nd pass. 35% was completed in 4 months)! So what have we been doing all this time?

You can also see an OGR-25 graph that shows in Dec 2000 it was >70% done. That was the 1st pass only, though.

So I'm still wondering how far along we are, and why we are still crunching if we were near completion by the end of 2000. There are relevant questions and answers in the FAQ about why there isn't a completion percentage on the stats pages, but it would be nice to have an update after more than 2 years.

Any thoughts peeps?

Lastest graphs (http://www1.distributed.net/~pstadt/ogr/). Not sure if this is the case, but looks like all the clients above v.460 are making "good" contributions to OGR's progress?

"There are a number of client versions earlier than .460 that provide the ability to process OGR, however due to known verification bugs in those earlier versions all work computed by them will be discarded by the servers." from the FAQ's.

[alpha]

Originally posted by Beyond
Lastest graphs (http://www1.distributed.net/~pstadt/ogr/).

Yeah, I did see those graphs but they don't really tell you anything other than how fast we're crunching.


Not sure if this is the case, but looks like all the clients above v.460 are making "good" contributions to OGR's progress?

"There are a number of client versions earlier than .460 that provide the ability to process OGR, however due to known verification bugs in those earlier versions all work computed by them will be discarded by the servers." from the FAQ's.

Interesting. But unless that FAQ entry is newer than Oct 2002, surely it is incorrect.

[alpha]

Some up to date news. (http://lists.distributed.net/hypermail/rc5.Feb2003/0029.html)

[Condor]

Originally posted by IronBits
:rotfl:
That damn psecky Phil on RC5-72 :bang:
and Condor on SOB trying to beat me to the 2TcEM spot :swear:
/me mumbles and wonders if he can get the HDTV decoder to crunch :)

I think you have gone off the deep end paranoid.:crazy: