Fast computer with 20 pending results

Evan
Evan
Joined: 23 Feb 05
Posts: 7
Credit: 117,469
RAC: 0

[quote[implies to me that the

[quote[implies to me that the 7 day deadline is really irrelevant: credit is given to all who have ever successfully returned a WU once some future gang of 3 confirms in its own 7 day window.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 4,979
Credit: 32,477,785,652
RAC: 31,413,263

> [quote[implies to me that

Message 6209 in response to message 6208

> [quote[implies to me that the 7 day deadline is really irrelevant: credit is
> given to all who have ever successfully returned a WU once some future gang of
> 3 confirms in its own 7 day window.
>

Cheers,
Gary.

Ken Vogt
Ken Vogt
Joined: 18 Jan 05
Posts: 41
Credit: 321,783
RAC: 0

> > I can assure both of you

Message 6210 in response to message 6209

>
> I can assure both of you that the above is most definitely the case. I've
> actually seen instances where it has take several weeks for the final third
> valid result to come in at which point everybody in the group gets their
> credit. You really have nothing to worry about.
>
Gary,

This is truly great. :) Perhaps it's only me, but it seems that two groups have been talking past each other here:

I/we ask about our work units not being sent out within x days, expiring without result; and we even see an occasional RSOD "No Reply."

Then you say "Don't worry, it will get sent out eventually and you *will* get credit."

And I/we think, "Yeah right, no way that can happen before deadline."

IOW, I/we think your answer is only in context of the oft-quoted "deadline," but in fact, your answer is to be taken literally. You are not being Pollyannas when you say Evan's 5 day old WU will clear, you *know* it will!

Now as I say, maybe I'm the only one with that impression, but I think that the several other threads here about slow validation may indicate there is widespread confusion on this point. And again, I may not have read this board closely enough and this point has been made clearly elsewhere, if so I am heartily sorry.

But I think with 10K or so new users just coming in to the "Where's my credit" stage, and I can't believe I'm saying this to eminent people whom I respect and admire, I think you ought to shout this fact out clearly, in the FAQ or better on the getting started page:

"Dudes! Don't worry about the frigging deadlines! When we get confirmation from three boxes, so that we're sure the result is sound scientifically, everybody who ever crunched the work unit successfully will get credit for it! Chill out!" :)

OK, it should be phrased in a more restrained fashion. :) It's just I think that if this message were clearly out there, you'd be having a lot fewer unhappy people now and especially in future.

I mean, hearing this changed Evan's view; enough said. :) Sorry, Evan, couldn't resist a poke. :)

Also the "Time reported and deadline" user page, which is pretty stark as written, could be modified to reflect this flexibility.

Offered in the spirit of constructive criticism; no disrespect intended to the team, quite the opposite in fact.

Edit after posting:
After a rereading of a few threads to see if I had indeed overlooked something, I saw the first post in this very thread from Dr Allen in a new light:

>Bottom line: counsel patience. If the user is returning successful results,
>they should get lots of credit. But it may take a week *or two* to start
>arriving.

I overlooked the "or two," which would imply there is no final 7 day limit. My only point is that this needs to be shouted from the rooftops.

Ken

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 4,979
Credit: 32,477,785,652
RAC: 31,413,263

> Now as I say, maybe I'm the

Message 6211 in response to message 6210

> Now as I say, maybe I'm the only one with that impression, but I think that
> the several other threads here about slow validation may indicate there is
> widespread confusion on this point. And again, I may not have read this board
> closely enough and this point has been made clearly elsewhere, if so I am
> heartily sorry.

Ken,
You are correct - there have been many threads expressing concerns at slow validation. In very many cases, there have been personal replies from Bruce Allen, very clearly and unambiguously stating that all you need is patience and you *will* get your reward. This has obviously taken a *lot* of effort on his part and has earned my respect and that is why I've been trying to assist with getting the message out, even though I'm just a simple participant like you and Evan. I'm mindful of five years plus experience with Seti classic where the silence of the admins was deafening by comparison. That experience tells me that the EAH boards have an unusually strong admin presence by comparison and I hope that will continue.

> But I think with 10K or so new users just coming in to the "Where's my credit"
> stage, and I can't believe I'm saying this to eminent people whom I respect
> and admire, I think you ought to shout this fact out clearly, in the FAQ or
> better on the getting started page:

But it *is* in the FAQ and it is in many, many posts. The FAQ is under rapid development with many messages per day by Bruce Allen saying "I'll put something in the FAQ about that", if you read them. I'm sure he is smart enough to realize that a single well thought out FAQ entry will save him a thousand similar replies to a thousand similar questions over the coming months. However this is all evolving so rapidly that it will take him time to put all those well thought out answers in place.

> "Dudes! Don't worry about the frigging deadlines! When we get confirmation
> from three boxes, so that we're sure the result is sound scientifically,
> everybody who ever crunched the work unit successfully will get credit for it!
> Chill out!" :)

Here is how it is phrased in the FAQ (Q.8 I think). Also check out the previous question and answer. How many times do you want it? :). :-

The credits are pending because two other machines have not yet completed the same work, and so it can't be validated yet. Don't worry: the BOINC scheduler will send that work to other machines (and keep trying if they don't do the work) until the same work has been done by several other machines. can typically take a week or ten days, or sometimes even a bit longer. Just be patient.

I guess he could add an extra sentence saying:-

The fact that you have pending credit proves that you have completed a result within the deadline so that you now don't have to worry at all about how long it takes for the required extra two valid results to appear.

How does that grab you? :). If every project supporter picked out an area of lack of information or weakness in the FAQ and drafted up and posted a suggested replacement or addition, I'm sure it would be a great help to the project Admins.

>
> OK, it should be phrased in a more restrained fashion. :) It's just I think
> that if this message were clearly out there, you'd be having a lot fewer
> unhappy people now and especially in future.

I actually think that this project is doing remarkably well in "getting the message out there" when you consider how young it is and how much has to be done behind the scenes to stop the edifice from collapsing under the dramatic weight of expansion that is going on. There's more than 1,000 new users every day at the moment, a figure likely to increase further if Seti keeps imploding.

> I mean, hearing this changed Evan's view; enough said. :) Sorry, Evan,
> couldn't resist a poke. :)
>
> Also the "Time reported and deadline" user page, which is pretty stark as
> written, could be modified to reflect this flexibility.

I presume you are referring to the page you get when you click the "view" link for your results. This page has a "Time Reported & Deadline" column with an "explain" link, as do other columns. I actually thought this was quite informative "as is" particularly with the ability to "drill down" into individual units of work to see exactly what was going on.

> Offered in the spirit of constructive criticism; no disrespect intended to the
> team, quite the opposite in fact.

I'm sure that nobody will take any offense at any of your comments. It's all very new for all of us and if we all adopt a "help out wherever possible" attitude, we will all be a very happy and productive community.

Cheers,
Gary.

Ken Vogt
Ken Vogt
Joined: 18 Jan 05
Posts: 41
Credit: 321,783
RAC: 0

Gary, I apologize if my

Message 6212 in response to message 6211

Gary,

I apologize if my language was excessive in the last post. I was so relieved that Evan's concerns had been resolved, I got giddy. :(

I concede all your points, in particular the item in the FAQ.

In several of my posts I have mentioned how grateful I am for the responsiveness of Dr Allen in particular.

I'd be happy to try to write a FAQ paragraph on the topic "Why are fast computers slow to get credit, and slow computers fast?", or any other topic you suggest. If I can use the phrase "Dudes!" in it. :)
>
>How many times do you want it? :). :-
>
I agree, it's been said many times, I've said it over at BABB, and quoted the responses from you , Dr Allen, and others there extensively.

But with the greatest respect, what has not been so clear until now, at least to me, is the issue of *deadline.*

In the drop down list you refer to, after seven days, if your WU has not been confirmed you get a line like:

1218119 354185 19 Feb 2005 14:43:42 UTC 26 Feb 2005 14:43:42 UTC Over No reply New 0.00 --- ---

The second date is in red; and your previous pending credit for the WU is wiped out to zero.

The stark screen I was referring to comes when you click "Time reported
or deadline [explain]" link in this detailed listing view. The URL is:

http://einstein.phys.uwm.edu/explain_state.php?field=result_time

and the text includes

"Not reported yet, deadline in the past; Deadline, shown in red."

If this page were even to add text such as "Although the deadline has passed, work will still be sent to more computers, and when confirmed, your credit earned will be restored," it would be a huge help.

In other words, again with respect, the deadline is not viewed by us as an individual deadline (for our computer to report), because, having successfully reported a WU, our pending credit is in fact wiped out for it when the deadline arrives.

Which is why I was so happy to hear explicitly from you, that even such wiped credit will be eventually restored.

Now I certainly agree that there have been a huge number of replies reassuring us that we will get all our credit eventually. A few, like the reply from Dr Allen I cited in my Edit in the last post, and the FAQ item you cited, do mention periods greater than 7 days, but it would certainly help if it could be explicitly stated along with the reassurances, that the "Deadline" mentioned so prominently in the user pages, is not in fact an "ultimate" deadline.

Your addition to the FAQ item, especially if it could also explain that the wiping of credit because other boxes fail to report within the deadline is not irreversibe, would indeed be a big help.

This is what I meant by us talking past each other: We worry about a "deadline;" but for you it's not an issue, because the deadline is individual; so your replies don't speak explicitly to the "deadline" which we see looming in the results section; we think we are being put off by the reply not speaking explicitly to this software deadline; you think we are being dense for not understanding clear statements of fact.

Both sides, I submit, are mostly right, but talking past one another.

I think users like Evan should be forgiven for not being able to take on board that a phrase like "[it] can typically take a week or ten days, or sometimes even a bit longer" is tantamount to an explicit voiding of the 7 day deadline.

Lastly, isn't the sheer fact that there have been so many questions on this issue a sign that it could be more clear?

I of course, as a suppposed team founder, have no excuse for not reading the FAQ item mentioned. For that, and for my tone in the last post, and in particular the word "Pollyanna," which was inexcusable, I apologize. I believe all in all I have tried to be constructive here, and if I have not, and have lost the respect of Dr Allen, I am devastated.

Ken

Darren
Darren
Joined: 18 Jan 05
Posts: 94
Credit: 69,632
RAC: 0

> In the drop down list you

Message 6213 in response to message 6212

> In the drop down list you refer to, after seven days, if your WU has not been
> confirmed you get a line like:
>
> 1218119 354185 19 Feb 2005 14:43:42 UTC 26 Feb 2005 14:43:42 UTC Over
> No reply New 0.00 --- ---
>
> The second date is in red; and your previous pending credit for the WU is
> wiped out to zero.

But your deadline is just that, it is your deadline - not everyone working that wu's deadline.

Here's one from my pendings (I'm computer 22531 on that list), and as you can see I'm still pending even though it's been 11 days since it was sent to me.

http://einsteinathome.org/workunit/357292

If you look at that example, you will see that it was sent to me on 20 Feb and sent to 3 other users on 23 Feb (so my due date was 27 Feb, while their due date was 2 Mar). Only one of those 3 returned it, so 2 of us have pending credit now and you'll see at the top of the list that 2 more wu were generated to replace the 2 that expired without being returned (and those wu are still unsent). If you click those unsent wu's, you will see that they were created on 2 and 3 March - so they were created after the deadline passed for those 2 users who didn't return it. At that point, even my original deadline had passed, but einstein still keeps the wu alive and keeps generating more to go out to get the required number back.

However, your pending credit should not be going to 0 (and your entry shouldn't be turning red) when your original deadline is passed as long as you returned your instance of it before that deadline. In the example above, you'll see that my pending credit - as well as the pending credit for the one other person who has thus far returned the unit - is still there. Only the ones that didn't return it go to 0.

Paul D. Buck
Paul D. Buck
Joined: 17 Jan 05
Posts: 754
Credit: 5,385,205
RAC: 0

In a new nutshell, If you

In a new nutshell,

If you return a result within the deadline, it will be eligable for credit. If there are problems with others returning work, the work will be re-issued up to the error limit set by the project until this error limit is reached or a Quorum of Results is formed.

The addition that needs to be made is that even if you beat YOUR deadline, if the result is not valid, then it will not earn credit.

If the work you returned is in fact valid, but a Quorum of Results cannot be formed before the error limit is reached you will not earn credit as the work unit will be invalidated.

So, it is not a simple linear process.

Few things in BOINC are as simple as they might first appear. SETI@Home and most of the first generation DC projects have FAQs with maybe 20 to 50 FAQ items. I have 480 something (If I remember right) and new ones are being added every week. And I don't duplicate project FAQ topics. So, BOINC is harder than the first generation projects and properly so ...

Ned Ludd
Ned Ludd
Joined: 9 Feb 05
Posts: 23
Credit: 56,045
RAC: 0

> In a new nutshell, > > If

Message 6215 in response to message 6214

> In a new nutshell,
>
> If you return a result within the deadline, it will be eligable for credit. If
> there are problems with others returning work, the work will be re-issued up
> to the error limit set by the project until this error limit is reached or a
> Quorum of Results is formed.
>
> The addition that needs to be made is that even if you beat YOUR deadline, if
> the result is not valid, then it will not earn credit.
>
> If the work you returned is in fact valid, but a Quorum of Results cannot be
> formed before the error limit is reached you will not earn credit as the work
> unit will be invalidated.
>
> So, it is not a simple linear process.
>
> Few things in BOINC are as simple as they might first appear. SETI@Home and
> most of the first generation DC projects have FAQs with maybe 20 to 50 FAQ
> items. I have 480 something (If I remember right) and new ones are being
> added every week. And I don't duplicate project FAQ topics. So, BOINC is
> harder than the first generation projects and properly so ...

One addition: if you return your work before a Quorum is formed, you'll get credit.

... and I think if you return work before it transitions off the BOINC database you'll probably get credit.



Ken Vogt
Ken Vogt
Joined: 18 Jan 05
Posts: 41
Credit: 321,783
RAC: 0

I do appreciate your replies.

Message 6216 in response to message 6215

I do appreciate your replies. Believe me, I understand *now* that the credit will come.

In my last post I mentioned I had a red deadline time out. This was also my error:

I had stupidly thought looking at BoincLogX that all my WUs had returned without error. But it turns out the WU in question, after closely comparing the "H1_" string, is not listed in BoincLogX at all, and so I "woke up" to the possiblity that it was a computation error on my box from the git go; in which case the red time out is entirely appropriate. Checking the details of the WU, which I either did not do or did not understand at the time, it is now obvious that that this WU did error out on my box. The 7 day deadline here was indeed an individual deadline, it passed, and no pending credit was ever withdrawn, because it had never been issued, all precisely as you have said.

I am very sorry to have made this mistake and caused you this aggravation, and so in addition to apolgizing for the intemperate language, I now withdraw unreservedly all comments made from Message ID 6758 onward, and apologize if I impugned your integrity or expertise. In short, I am a doofus, who needlessly hurts good people; I'm really, really sorry.

In future I will come here, if I may be still permitted to, only to post specific questions for BABB members who don't have time or desire to come here themselves.

Ken

StarCharter
StarCharter
Joined: 19 Feb 05
Posts: 59
Credit: 641,079
RAC: 0

> I am very sorry to have

Message 6217 in response to message 6216

> I am very sorry to have made this mistake and caused you this aggravation, and
> so in addition to apolgizing for the intemperate language, I now withdraw
> unreservedly all comments made from Message ID 6758 onward, and apologize if I
> impugned your integrity or expertise. In short, I am a doofus, who needlessly
> hurts good people; I'm really, really sorry.
>
> In future I will come here, if I may be still permitted to, only to post
> specific questions for BABB members who don't have time or desire to come here
> themselves.

Ken,
Don't be so hard on yourself! We all make mistakes and many of us stick to what we think long after it would have been better to abandon it. I can't speak for anyone else but you haven't hurt me in any way and I'll always be glad to see you in here.

There are two secrets to life: 1) Don't tell everything you know...

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.