I would suggest setting 'no new tasks' to almost all your projects and just pick one or two projects to run for now. You can switch running projects in a few weeks or a month.

Or just reduce the cache to 0 days + 0.01 additional days.
Change 2 values & no more missed deadlines & projects being blocked from work due to existing workload. And no need to change things again in the future. Other settings should be changed to improve things further- but the cache settings are the main ones causing the biggest issues.

---

Grant
Darwin NT

PrimeGrid is showing on your computer info page here at Rosetta. Following the link to your computer and tasks at PrimeGrid shows contact Feb 16th and a task 'timed out and no response'

https://boinc.bakerlab.org/rosetta/show_user.php?userid=2400235

http://www.primegrid.com/hosts_user.php?userid=1646462
http://www.primegrid.com/results.php?hostid=1323608

I would suggest setting 'no new tasks' to almost all your projects and just pick one or two projects to run for now. You can switch running projects in a few weeks or a month.

It's simple, Rosetta does not generate a lot of work consistently. There are over 1 million users on this project and 4 million computers. Out of that in the last 24-48 hours or whatever they call 'recent' there were 14,977 users with 30,100 computers and I just came back to look at whats going on here since I stepped away and saw that there was something like 100,000 tasks. So if you divide that total by computers with multiple cores you can see that batch of work was gone in a heartbeat. I came back just now 24 hours later and the work is all being processed and there is no new work.

You can stay attached to this project, but the work will be in spurts and not steady.
They do most of their work inhouse now on a AI system.

First, I reduced my projects to four by detaching from several.

Second, I have now successfully completed two tasks from Rosetta, so things are looking up.

Now everything is running well. I have positive results in all four projects.

Thanks for thinking of me.

---

Please check out my blog
http://sciencesprings.wordpress.com
http://facebook.com/sciencesprings

First, I reduced my projects to four by detaching from several.

Which probably won't help much (if at all), as the main cause of your problems are your computation settings. Which you refuse to do anything about.
Once again, you choose to do something unnecessary, and ignore that which would actually help.

Second, I have now successfully completed two tasks from Rosetta, so things are looking up.

Now everything is running well.

Having Tasks time out due to not returning them in time is not running well (as is taking 3.5 hrs to do 3hrs of work). When those 2 Tasks were 50% of the work you had, it's actually running extremely poorly by any reasonable standard.

---

Grant
Darwin NT

Sorry, the other projects are running too well to allow Rosetta to force them out. If Rosetta cannot get enough run time, so be it. But two tasks were completed on time.

Thanks for your interest and comments.

---

Please check out my blog
http://sciencesprings.wordpress.com
http://facebook.com/sciencesprings

Richard,

My configuration can handle 10 projects.
I keep a small queue for work. .75 for primary and .75 for extra.
Anything more than that and you don't return tasks on time.
I assign 95% of my memory resources to BOINC and 15/16 threads.
I run a 8 core 16 thread processor and 2 GPU cards.
I change tasks automatically every hour.

I don't know how much work you are keeping, but you should not keep more 1.5- 2 days in total of work. Less is better. You get new work faster from the projects and you do not run out of time for credit if there is a short deadline.

My kind of configuration allows Rosetta to get its time in along with other projects.
Key point of all this is keep your work units value low and let BOINC balance how much work you do from each project. It might download a bunch of Rosetta work first to get the credits up and then once it has reached its happy point it will settle down and balance your work load with the other projects. If you go messing around with stuff to much, then BOINC has no idea what to do anymore. Try .50 in both fields for the amount of work. Then just let BOINC work on the tasks you have in queue right now and let it get to 1 day of total work. If your projects are happy with this, then change to .75 in both or some mix that equals 1.5 days of work. You can try 2 days and see what happens, but I think that is to much work and will create a backlog with deadline missed abort errors.

This is just my personal experience with BOINC and Rosetta and all the others.

Thanks for your comments. I am running only four projects, pretty much all on default configurations. BOINC sets the queues. Rosetta has twice succeeded in completing tasks on time, in both cases two tasks. I do not interfere in BOINC's process. I have learned, painfully, the less I touch the better off I am.

I am running Win 11 23H2. My Intel i7 processor has eight cores and my GPU is Nvidia compliant.

Thanks again.

---

Please check out my blog
http://sciencesprings.wordpress.com
http://facebook.com/sciencesprings

Sorry, the other projects are running too well to allow Rosetta to force them out.

Complete and utter nonsense.
One of your other projects has even more missed deadlines than Rosetta.

24 missed deadlines out of 34 total. 70% failure rate. Beyond abysmal.

---

Grant
Darwin NT

I am running only four projects, pretty much all on default configurations.

Once again, not true.
As i have pointed out time and time again, the majority of your settings are not the defaults, and that just changing your cache to 0.1 days and 0.01 additional days would resolve the missed deadline issues and the difficulty in getting Rosetta work. Yet you continue to do as you always do- Make statements that are factually not correct, ignore advice you have asked for, then go and change something else entirely different to what has been suggested (in this instance removing some projects).

Rosetta has twice succeeded in completing tasks on time, in both cases two tasks.

And two that missed the deadlines- 50% failure rate. Abysmal.

I do not interfere in BOINC's process.

Yet the evidence shows that you have, and this thread shows you continue to do so.

I have learned, painfully, the less I touch the better off I am.

And yet, once again you just go ahead & do something that wasn't suggested, as it wasn't necessary and will likely have little if any impact.

---

Grant
Darwin NT

Richard,

Grant and I are old hands at this stuff.

I will suggest this, you have 26 Roesetta tasks that take 6+ hours each to finish on a 3 day deadline.
That gives you very little room to do any other work.
For now, set EVERYTHING to 'No new tasks' in the project tab and empty your queue of all projects and all tasks.
Change your days of work to .25 in both boxes. That is a half day in total.
When your queue is empty then accept new work.

I see your on LHC like me, but yet you have very little recent credit.
I don't have time to see whats going on there, but you should be getting more recent credit from that project unless your just waiting on ATLAS.

Based on your RAC you should dump these projects:
PrimeGrid 942,098 4 18 Nov 2023
LHC@home 3,930 1 13 Nov 2023
GPUGRID 17,816,440 0 8 Feb 2012
yoyo@home 76,333 0

And WCG...you might as well dump that. They can't get their act together.

Then you are down to:
Einstein@Home 1,523,996 46,740 9 Nov 2023
NumberFields@home 136,160 2,838 30 Nov 2023
MilkyWay@home 23,784 482 9 Nov 2023
Asteroids@home 134,897 309 7 Nov 2023


And Rosetta.

With this done you have a clean set up of BOINC.
But again... .5 days in each field. This should give you a nice fast turn around and improve your chances of getting credit.

Follow these recommendations and you should see an improvement.

BTW... .1 and .01? what the *%*%* is that nonsense. At minimum .25 and .25 for a half day queue. or .25 and 0 for a 1/4 day. But honestly....stop messing around with those values. .25 or .50 or some combination of the two for a quarter or a half or three quarters of a day of queued work. But first, NO NEW WORK!

BTW... .1 and .01? what the *%*%* is that nonsense. At minimum .25 and .25 for a half day queue. or .25 and 0 for a 1/4 day. But honestly....stop messing around with those values. .25 or .50 or some combination of the two for a quarter or a half or three quarters of a day of queued work. But first, NO NEW WORK!

Once again you've got it wrong.

Set the cache to an appropriate value, which for more than one project is 0. There is no need for a 3/4 day, 1/2 day or 1/4 day cache- it is completely, totally and utterly redundant & pointless unless every single one of the projects you are attached to has intermittent work availability or server up time. Since that isn't the case, there is absolutely no need for a cache of work.
And that alone will stop new work from coming until the system is ready for it. There is absolutely no need to remove projects, no need to set No new work on one or more projects, then re-enable it again later on. There is absolutely no need to fart-arse around with such things when setting the cache to 0.1 and 0.01 will do it for you, and stop the issue from re-occurring in the future.

Just set the cache to a reasonable level- and for multiple projects that is no cache, which 0.1 & 0.01 provides, while still allowing enough work to download so that when one Task finishes, another is ready to run & you don't have to wait for it to download before it can start. Any more fiddling around than those 2 values is unnecessary.

---

Grant
Darwin NT

Thanks Grant and Greg, but I am not touching anything. Rosetta succeeded again with good results. Yes, there is a lot running, I hope some good comes of it, but I will leave it up to BOINC to sort it out.

Wow!! I even got the Rosetta screensaver. It is like the good old days of my first adventures with BOINC a number of years ago when I racked up 36,000,000 credits. In those days, there was none of this fiddling around. One just attached to whatever projects one could, set the CPU to something like 50% in BOINC, not in any projects and let things run.

So, now, whatever happens happens.

---

Please check out my blog
http://sciencesprings.wordpress.com
http://facebook.com/sciencesprings

BTW... .1 and .01? what the *%*%* is that nonsense. At minimum .25 and .25 for a half day queue. or .25 and 0 for a 1/4 day. But honestly....stop messing around with those values. .25 or .50 or some combination of the two for a quarter or a half or three quarters of a day of queued work. But first, NO NEW WORK!

Once again you've got it wrong.

Set the cache to an appropriate value, which for more than one project is 0. There is no need for a 3/4 day, 1/2 day or 1/4 day cache- it is completely, totally and utterly redundant & pointless unless every single one of the projects you are attached to has intermittent work availability or server up time. Since that isn't the case, there is absolutely no need for a cache of work.
And that alone will stop new work from coming until the system is ready for it. There is absolutely no need to remove projects, no need to set No new work on one or more projects, then re-enable it again later on. There is absolutely no need to fart-arse around with such things when setting the cache to 0.1 and 0.01 will do it for you, and stop the issue from re-occurring in the future.

Just set the cache to a reasonable level- and for multiple projects that is no cache, which 0.1 & 0.01 provides, while still allowing enough work to download so that when one Task finishes, another is ready to run & you don't have to wait for it to download before it can start. Any more fiddling around than those 2 values is unnecessary.

Geees you are a sour old man...this setting works perfect for me. However I will try the 0 setting.
If the cache is not needed then why did BOINC put it in? There is obviously some benefit to it. BOINC developers don't put stuff in for no reason.

I get rid of useless projects because my system goes looking for work there and if there is none then why keep them? Especially projects that are dead. Perhaps though that is just stuff he was attached to in the past. I didn't go digging around.

Anyway..hes set in his ways...so whatever...I'm gone.

Geees you are a sour old man...

Pointing out the same thing to the same people over & over again in just a single thread does have that effect.

If the cache is not needed then why did BOINC put it in? There is obviously some benefit to it. BOINC developers don't put stuff in for no reason.

Caches are a hangover from dialup days where people didn't have always on internet access and had to pay the call connection fee every time they connected to the internet. It allowed them to get work for several hours, or several days, or even a week, to avoid having to connect regularly to return & get new work, and having to pay each & every time they had to initiate a connection if they already weren't online at that time. It's also from the time when there was just one project, and the supply of work was at best irregular.


And today, as i have pointed out already in this thread- even if you have just one project, if it is reliable with it's servers being up and work availability, then there is no need for a cache. If it's not, then a cache will keep your system busy during those periods the project is down.

But when you have more than one project, especially so when you're talking more than 3, the odds of all projects having issues all at the same time is small to non-existent, so there is absolutely no need for a cache in such cases. If one or more projects goes down or runs out of work, the other projects will keep your system busy. When they come back up, your Resource Share settings will once again be honoured as they get work from those projects that were previously down, and do less work from the one(s) that weren't, until things have balanced out again.

---

Grant
Darwin NT

I have overnight again had great success at Rosetta,

---

Please check out my blog
http://sciencesprings.wordpress.com
http://facebook.com/sciencesprings

Geees you are a sour old man...

Pointing out the same thing to the same people over & over again in just a single thread does have that effect.

I haven't been around for ages - I'm so busy - but checking in and reading these threads are enormously funny at a distance, so please carry on.

It reminds me of someone years ago who kept a 10 day cache, running projects with like 3 or 7 day turnarounds and wondering why he wasn't getting any credit.
Brilliant stuff. Nearly killed me, but brilliant.

And then the Seti mob arrived. In hindsight, also glorious.

I can laugh now. The passive aggressive posts were through the roof.
Some people can't be helped.

My advice, for what little it's ever worth or has ever been taken up, is to analyse the overall problem offline, then make a single <tiny> suggestion online.
Something seemingly harmless that will receive least pushback, with the mere suggestion to leave it like that for a few days, then ask to feed back how things are going.

If it works, it works. If it's rejected, making more suggestions won't get any further.

And don't make the suggestion extreme. Greg's suggestions are fine (clearly). Richard's <clearly> aren't.
Pick one. Get success/do no harm/gain trust then move onto the next little tweak.

The idea is for everyone to be successful by their own standards.

---

While you guys can discount my approach and everything I have said, things are going well now with all four projects to which I am attached, especially Rosetta which continues to return positive results.

---

Please check out my blog
http://sciencesprings.wordpress.com
http://facebook.com/sciencesprings

While you guys can discount my approach and everything I have said, things are going well now with all four projects to which I am attached, especially Rosetta which continues to return positive results.

Well then why come here and ask detailed questions when you don't follow his information?
Look, in the end if everything works ok, then just leave it alone.
But Grant has left very specific instructions to follow if things don't work.

Before you ask questions about no work, look at the server status page. If the ready to send is 0 or 1 then there is no new work available at the time you looked. Check again in 12-24hrs. If you still have no work, then either your queue is full with other work or too small. The log will tell you. Requesting no work or requesting no work CPU cache (or whatever the message is) means you queue is full. Again the decimal figure needs to be higher or 0 if you want more work, otherwise just wait.

While you guys can discount my approach and everything I have said, things are going well now with all four projects to which I am attached, especially Rosetta which continues to return positive results.

Hi John,
I'm glad you're Projects and Tasks are running to your satisfaction. That's a big positive and solved your primary reason for starting this thread.
Before you weren't able to get or complete any Rosetta tasks and now you can, so that's a win by anyone's measure.

The guys here are a bit snarky, but only because they're well-meaning. Genuinely.
To tell the truth, I've been even worse than them in my past (full disclosure) but I've eased back over the last year.
Part of their reason is we can all see you have a decent PC and they're keen for you to get the very most out of it.

Of course, there's no need for you to get the <very most> out of it as long as you're comfortable, but some of their suggestions really will be able to improve <all> your projects all at the same time - not improving one at the expense of another, which is what seems to be your biggest concern - and a perfectly understandable and reasonable one.

Grant certainly does know what he's talking about, but his approach is to correct everything all at once and to tune everything to the finest margin and that's what's concerning you istm, which doesn't surprise me as I don't think I've ever tuned my PC to the level he runs at. Greg is more pragmatic, but his advice is no less good. With the right assurance you can trust what either are advising (or any combination of the two).

Now, I know you're expecting me to come up with some suggestion of my own, but before I do any of that, there's something you've been saying that's not right - not right at all.

There's been some discussion as to whether you're running default settings for each project.
Imo, while it's certainly true some of your settings are <not> at their default
1) Largely I don't think that's the point
2) To some extent it doesn't even matter
So don't get bogged down by it. Instead...

Projects run <completely> independently of each other. Default settings for one project bear no relation to any other project.
If you're only running a single Project, the default will generally (but not always) be fine for that project.

But when you're running more than one project, project defaults may be in direct (and disastrous) conflict with each other.
So saying you're only running defaults for multiple projects (8 or 10 in your case) is not a get-out clause.

As evidence for what I'm saying, Seti closed down a few years ago and a lot of the most expert Boinc users with the most highly spec'd machines joined Rosetta with their highly-tuned project settings and pretty much took the whole site down. Not because they didn't know what they were doing, but because the finely-tuned settings they had for Seti were completely inappropriate for this project. And you can imagine how well many of them took it when they were advised how to make best use of the project here. Believe me, it wasn't pretty, but we got through it... (eventually, kicking and screaming). So none of the suggestions being made to you are condescending or doubting your abilities, experience or expertise. It's really not like that.

Imo this partly ties in with why you're now running successfully, when you weren't at the start. You said you removed a few Projects, which would have removed a few of those conflicts and at the same time would've reduced the cumulative demands on Boinc managing tasks from all of them at the same time with the cache of tasks you request. Grant is quite right that you didn't need to remove projects if that's what you originally wanted to run, but nevertheless, removing some has relieved a lot of pressure imo.

Which brings me onto another factor.
Boinc has settings separate from each Project requirement.
And from what I can gather, it hasn't been Project settings that have been the cause of the bulk of your issues, but Boinc settings in relation to them.
And it's Boinc settings, not Project settings, default or not, that people have been rightly advising you about.

In the context of the large number of Projects you originally wanted to run, Boinc settings need to be a bit of a compromise between the requirements and limitations of all your projects, which can be quite the balancing act at times. The reason I think you started this thread is because you got your Boinc settings just the wrong side of the line of getting your Projects to run nicely together. And if you're saying they're running ok now, it's only because of the Projects you removed and not because you've got Rosetta running well, because objectively they are running really, really badly (I can't emphasise this enough - they're running terribly) . It's just that running badly is better than not running at all.

Now, I'm not experienced with a lot of Projects myself. I only run Rosetta and World Community Grid, so I'm unfamiliar with the requirements and limitations of all the others. But I note the user "Link" points to a problem with Project Milky Way which gave you a task problem the other day, so take advice from him on that one, but it is a clue you've still got issues under the surface, even if you do have Rosetta running (limping) along at the moment.

After all that preamble, which I hope you take on board in the spirit given, I'm going to suggest a reason for the problem you originally came here for - the lack of Rosetta tasks when tasks were available to download, even while you got tasks from other projects. This will not change a single thing about how your tasks run, so if you're running successfully now, you'll continue to do so with absolutely no change in that.

In your Boinc Manager settings (from Options/Computing Preferences/Computing tab, right at the bottom) you have set:

Store at least 0.00 days of work
Store up to an additional 3.00 days of work

For the Rosetta Project (and I'm guessing none of your other projects are the same or less) the deadline is set at 3 days after you download tasks.
Which means that when Boinc calculates the total runtime of the tasks you 've downloaded from all your projects across all your processor cores, if it totals 2.99 days it's got room to call a task down.
So, in theory at least if not in practise, 2.99 days after you download it, it will start running. And 0.01 days after that it'll have missed its 3-day deadline because it takes several hours to run the task.
It goes without saying, for Rosetta specifically, this is not good. It's literally planning to fail.

So while your Rosetta tasks are currently running quite soon after downloading them <at the moment> once Boinc has settled down with all your projects, you've still got a problem waiting to hit you.

Grant's extreme suggestion is you should have these two settings at 0.00 and 0.01, while Greg has his set at 0.75 each.
The most important thing is the two figures combined have to be less than Rosetta's deadline plus an allowance for the runtime of the task itself (default is 8hrs runtime - 0.33 days).
And I think there's a general view that the first number shouldn't be zero (not sure where this comes from tbh, but it does make sense).

My personal preference is much closer to Greg's. I actually use 0.5 and 1.0 for a total of 1.5 days (same as Greg) towards a deadline of 3 days, which seems a nice compromise figure, but pick your own for a total of 1.5days.

What I think this will do for you is give Boinc lots of room to download Rosetta tasks but also other tasks from other projects as and when they come due.
When you have a lot of projects, like you do, it may be advisable to reduce that total further to 1.0 or even lower.
This is Grant's view - the lower the total, the more scope Boinc will have to manage tasks for all your other preferred projects and that's exactly why he uses such extremely low numbers.
It's also why he says you didn't need to remove all those projects - and why he complained when you did. It did turn out helping you but at the expense of removing Projects at some previous time you wanted to contribute to.
So while Greg and I are happy with 1.5 days total, if you want to contribute to more projects, knock it down further. 0.1 and 0.5 may suit your original preference for lots of projects.

And to emphasise again, this won't change how your tasks run when they're running. Just Boinc's ability to make tasks available from all your projects when they have them to offer you (which Rosetta is a bit dodgy with in recent months and years).

This also won't make your tasks run better - they really are running very, very badly - but let's do one thing at a time and allow you some confidence that suggestions here are good (and they have been good, in spite of your doubts).
There's no substitute for confidence and reassurance. No-one's trying to lead you astray here.

---

While you guys can discount my approach and everything I have said, things are going well now with all four projects to which I am attached, especially Rosetta which continues to return positive results.

This also won't make your tasks run better - they really are running very, very badly - but let's do one thing at a time and allow you some confidence that suggestions here are good (and they have been good, in spite of your doubts).
There's no substitute for confidence and reassurance. No-one's trying to lead you astray here.

Despite no reply here, I can see lots of tasks are running and they're running a little less bad than before, so you've taken some suggestions on-board. Great.
I think you were completing ~10,800secs of CPU work in ~30,000secs before. Now it's ~10,800secs of CPU work in 24-27,000secs.
Better, but still objectively terrible.

If you're still reading, here's the next list of suggestions:

Boinc settings - under Options/Computing Preferences
Think of these as Boinc restrictions from the Project defaults.

First the two easy ones:
4th tab - Daily Schedules
All options should be unchecked and all fields cleared

2nd tab - Network
All options should be unchecked and all fields cleared

I expect this is what you've already got set-up, so no problem here

1st tab - Computing
Again note: these are Boinc restrictions from the Project defaults

All checkboxes should be unticked and any field on that same line should be cleared - delete any figure you find there. You don't want any restrictions from project defaults.
Except the last checkbox called "Leave non-GPU tasks in memory while suspended"
This is a quirk at Rosetta that came up probably 10yrs ago due (I think) to a checkpointing issue at the time and it's worth keeping ticked just in case it recurs.

These are <very> minor changes (if they change anything at all) that will barely make any difference to the vast majority of users, but it depends how you personally use your PC.
It's just best to get them out of the way in order to prepare for the next idea for a change, which will start to make more of a difference.

I'll leave it 3-4 more days and see if either of us notices any performance change at all in your reported tasks.

Final note: these minor changes will slightly improve performance on <all> your project tasks, not just Rosetta, so even if Rosetta tasks run out (quite likely tbh) you'll get some benefit.

---