Tom Coradeschi - for #4 - choose some other projects to keep your system busy while waiting for non VM tasks (if that is all you want)

What do you want to do? Math, Astro, Health?

Pretty sure this will have been answered before but I didn't find it on a quick search. With nothing available from my normal projects I downloaded 16 tasks to run on my Ryzen7 All are showing

Sun 27 Feb 2022 11:26:55 GMT | Rosetta@home | Task aaam-ABU-PHE_pp-mNMABU-AMACBEN2_13_2375631_2_0 postponed for 86400 seconds: Communication with VM Hypervisor failed.

Ubuntu 21.10, BOINC7.19.0 (Built from source about two weeks ago.) VB6.1 installed.

Queued job is not so important as unsent tasks.
That is what tells us what is coming really soon.

No, it is the opposite. Queued jobs are the total they have (that they are telling us about at any rate).
The unsent jobs are just the work buffer. If they have anything in the queue, then the buffer gets filled.

When the queued jobs go to zero, that is when we run out.
Though if you can't do the pythons, then you probably want to keep an eye on that part. I do both.
But it would be better if they showed the difference in the queue. That is probably what you are referring to.

---

Tom Coradeschi - for #4 - choose some other projects to keep your system busy while waiting for non VM tasks (if that is all you want)

What do you want to do? Math, Astro, Health?

Yep - essentially what I was thinking with #3:-)

Well again, what are you looking for?

Lots of possibilities out there, a new one that has been mentioned that is not Vbox related is gene@home.
LHC@home for physics
Einstein for Gravity waves from black holes
SiDock (health related)
If you want to work your GPU's hard run Prime Grid.

Outside of BOINC some of us run folding at home.

Outside of BOINC some of us run folding at home.

A lot of people apparently avoid Folding because it uses its own client software. But I think it is probably the most innovative and useful project around.
It used to share that status with Rosetta in my estimation, but it is hard to justify that these days (for the work that they send to us at any rate).

I run both, usually using Folding on the GPU and BOINC on the CPU of any given machine. You just delete the CPU slot in Folding for that, and then reserve a core in BOINC for Folding on the GPU.
In fact, increasingly I am running the CPUs on Folding too. They don't get as many points as the GPU, but it is somewhat different science on smaller molecules (peptides) that appears to be quite important.

---

Outside of BOINC some of us run folding at home.

A lot of people apparently avoid Folding because it uses its own client software. But I think it is probably the most innovative and useful project around.
It used to share that status with Rosetta in my estimation, but it is hard to justify that these days (for the work that they send to us at any rate).

I run both, usually using Folding on the GPU and BOINC on the CPU of any given machine. You just delete the CPU slot in Folding for that, and then reserve a core in BOINC for Folding on the GPU.
In fact, increasingly I am running the CPUs on Folding too. They don't get as many points as the GPU, but it is somewhat different science on smaller molecules (peptides) that appears to be quite important.

It's not about points to me. Its about advancing the science with the hope they can use it to fight whatever they are looking at. You don't get anything for points, just a status symbol among BOINC users.

Queued job is not so important as unsent tasks.
That is what tells us what is coming really soon.

No, it is the opposite. Queued jobs are the total they have (that they are telling us about at any rate).
The unsent jobs are just the work buffer. If they have anything in the queue, then the buffer gets filled.

When the queued jobs go to zero, that is when we run out.
Though if you can't do the pythons, then you probably want to keep an eye on that part. I do both.
But it would be better if they showed the difference in the queue. That is probably what you are referring to.

But is the queue shown on the page the real queue?
I would love to see the breakdown by task type vs some number.

It's not about points to me. Its about advancing the science with the hope they can use it to fight whatever they are looking at. You don't get anything for points, just a status symbol among BOINC users.

Certainly. But it is a little more subtle on Folding, since both GPUs and CPUs do the same sort of thing, molecular dynamics (MD). So it is often thought that since the GPUs are getting more points, they are more valuable. But that is not quite how it works, since they are doing different types of work on different types of molecules. It helps to see the task descriptions, which the researcher provides for every work unit at Folding. That is quite different than here, to make an obvious point.

---

This is their best hope.
https://boinc.bakerlab.org/rosetta/forum_thread.php?id=14930&postid=105000#105000
But they have to have someone on the job to do it.

Who has the connection to the team?
Someone in touch with DEK?
He is the tech guru.

I have passed on those details from a PM sent to me.
The info was given in a very clear way, but I'm not certain if it's been acted upon or not.
I'll try to find my email and give a little nudge.

---

But is the queue shown on the page the real queue?

Yes.

I would love to see the breakdown by task type vs some number.

Computing, Server status, Tasks by application. Unsent numbers.

But as far as the main page Total queued jobs is concerned, you just have to remember roughly what that number was before any Rosetta 4.20 are sent out to have an idea of just how many Rosetta 4.20 Tasks there are queued up to go out (when there are some to actually go out).

---

Grant
Darwin NT

I have passed on those details from a PM sent to me.
The info was given in a very clear way, but I'm not certain if it's been acted upon or not.
I'll try to find my email and give a little nudge.

Well done. My little help: i wrote a message about "multiattach" on IPD, R@H, RosettaCommons twitter accounts.

---

I have passed on those details from a PM sent to me.
The info was given in a very clear way, but I'm not certain if it's been acted upon or not.
I'll try to find my email and give a little nudge.

Well done. My little help: I wrote a message about "multiattach" on IPD, R@H, RosettaCommons twitter accounts.

I would've sent the message of 4th February, having been sent it on 3rd Feb...
...it's not looking good, is it.
All my VB tasks are failing at the download stage atm. I'm running nothing at all for anyone :(

---

It's not about points to me. Its about advancing the science with the hope they can use it to fight whatever they are looking at. You don't get anything for points, just a status symbol among BOINC users.

Certainly. But it is a little more subtle on Folding, since both GPUs and CPUs do the same sort of thing, molecular dynamics (MD). So it is often thought that since the GPUs are getting more points, they are more valuable. But that is not quite how it works, since they are doing different types of work on different types of molecules. It helps to see the task descriptions, which the researcher provides for every work unit at Folding. That is quite different than here, to make an obvious point.

The task descriptions are very nice.
You understand better what you are crunching.
RAH could learn from that.
In the old days a grad student on occasion would write about what they were putting on the system.
That died a long time ago when everyone disappeared.

But is the queue shown on the page the real queue?

Yes.

I would love to see the breakdown by task type vs some number.

Computing, Server status, Tasks by application. Unsent numbers.

But as far as the main page Total queued jobs is concerned, you just have to remember roughly what that number was before any Rosetta 4.20 are sent out to have an idea of just how many Rosetta 4.20 Tasks there are queued up to go out (when there are some to actually go out).

If they have a million then are they mostly python or a mix with 4.2 and then why isn't 4.2 kept filled up?
Python is down to 4,xxx but there are a million in queue. So it really doesn't make any sense.

If they have a million then are they mostly python or a mix with 4.2

Like i said, you have to keep track of how many Pythons there are (Total queued jobs) when there are no Rosetta 4.20 Tasks available (Unsent number in Tasks by application).
If there are no Unset Rosetta 4.20 Tasks, then all of the Total queued jobs are Python. If the Total queued jobs number suddenly increases, and there are now Unset Rosetta 4.20 work, then that increased number is the number of Rosetta 4.20 jobs in the Total Queued jobs.

But if there aren't any Rosetta 4.20 Tasks in the Unset numbers, then that increase in Total queued jobs numbers means they are Python Tasks and you need to remember that new number to figure out how many Rosetta 4.20 Tasks are in the Total queued jobs number the next time it increases and it's caused by a batch of Rosetta 4.20 Tasks and not more Python Tasks.

and then why isn't 4.2 kept filled up?

You need to ask the project why they aren't sending out any Rosetta 4.20 work, and why so little on the extremely rare occasions when they do.

Python is down to 4,xxx but there are a million in queue. So it really doesn't make any sense.

Think of the Unsent number as what is queued up in the feeder ready to go to. The Total queued jobs is that number, plus all the other Tasks ready to go, that won't fit in the feeder.
That's why you can have 20 million in the Total queued jobs, and only a few thousand in the Unset.

---

Grant
Darwin NT

Python is down to 4,xxx but there are a million in queue. So it really doesn't make any sense.

Think of the Unsent number as what is queued up in the feeder ready to go to. The Total queued jobs is that number, plus all the other Tasks ready to go, that won't fit in the feeder.
That's why you can have 20 million in the Total queued jobs, and only a few thousand in the Unset.

Or think of it the other way, the work generators dump the tasks into the total queue and there are a couple of daemons running around checking whether the unspent queue each is responsible for is full (5,000 for Python, 27,000 for 4.20) and topping it up as necessary. There could be many reasons why they choose to separate the storage used by the work generators from that used by the task delivery function.

[snip]

and then why isn't 4.2 kept filled up?

You need to ask the project why they aren't sending out any Rosetta 4.20 work, and why so little on the extremely rare occasions when they do.

[snip]

But how do we ask the project anything and get an answer?

[snip]

and then why isn't 4.2 kept filled up?

You need to ask the project why they aren't sending out any Rosetta 4.20 work, and why so little on the extremely rare occasions when they do.

[snip]

But how do we ask the project anything and get an answer?

You don't because they don't answer.
Only one person here has a "in" with the project and it seems they ignore that person as well sometimes.
The project is what it is...we have to come up with the answers to any problems.
We are beginning to feel like leftovers after they got their neural network up and running.
That is when everyone disappeared or shortly after that.

Found out I can't run 15 pythons on a full throttle system and expect any sort of reasonable response time.
I have never used run based on preferences until now. Take back some resources when I am using the computer.