Question on deadlines; arbitrary or real?

Michael Tangen
Michael Tangen
Joined: 11 Mar 16
Posts: 2
Credit: 729384
RAC: 0
Topic 198631

So I've got a question on deadlines for all of the projects that my machine is running. Some of the projects that my machine has been working on are past their deadline. Are these deadlines very real (as in the data is useless after the deadline and won't be used) or are they a relatively arbitrary date? I'm just wondering if it's more productive to kill those projects that are behind or let them finish out anyway?

Could someone speak to this topic and give more information about what is behind the deadlines and whether or not they're really important for us to be concerned about (and going over)?

Runner • cyclist • beer and brewing geek • fascinated with space and time

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2139
Credit: 2752977655
RAC: 1378969

Question on deadlines; arbitrary or real?

That answer, as so often, is "it depends".

The first thing to check is probably the status of the tasks as displayed on this website:

All tasks for computer 12214507.

None of the older ones are visible - that means that since the deadline, another copy has been sent out to a different user, the result has been returned, the checks for accuracy have been completed, and the records have been purged (to save space in the database).

Once a workunit has reached that stage, there's no point in continuing to crunch it - the scientific results won't be used (they've got them already), and you won't get any credit. Abort the oldest five, but keep the three new ones.

The deadlines aren't absolute: if you are just a little bit late, you may still return it earlier than your replacement. In that case, your results will be welcomed and credited.

Some other projects are more 'absolute': if they don't have the 'checking' process of validation by a second user, they may cut you off as soon as the deadline passes. You might need to ask at those other projects too.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109409114541
RAC: 35172864

Hi Michael, Welcome to the

Hi Michael,
Welcome to the Einstein project.

Everything that Richard mentions is correct but I'd like to add a few details. In case you think "it depends" might imply a fair bit of extra time, let me spell out exactly what happens. Immediately after the deadline is reached with your task not returned, the scheduler starts the process of creating an extra copy and then sending that copy out to a further host. The delay in doing that might be as short as a few seconds up to a few minutes, or, depending on the particular science run the task belongs to, could even be hours or longer. Once the task has been reissued, there will be a further time delay for the recipient host to crunch the work and then return it. Immediately after the extra copy is returned and if it happens to be validated, your copy is of absolutely no further use.

If you have to make these sorts of decisions at any future time, your first step is to look at your tasks list - exactly what Richard provided the link for. You find this page for yourself by going to your account page on the website and clicking the 'view computers' link - in your case just one computer. You then click the 'tasks' link for that host that you find under the "Computer ID" heading (leftmost column). If you have expired tasks in your list, they will show there highlighted in red initially. For you it's already way past that stage because those tasks are already removed, as Richard mentioned.

I want to comment on what you could do if any were still showing. For any such task, click on the numerical workunit ID for that task that you find under the "Workunit" heading. This will show you the status of all copies of the task that have been created. If there are 2 returned tasks and the workunit is already completed, it's too late. If there is one or less returned tasks, there is still a chance. The only way to make any sort of assessment of what that chance might be, would be to look at the 'details' for each host that still has a copy to be returned. Yours will be one such host but there should be at least one other.

You can drill down into the 'details' for that other host and find when the task was issued and what the average turn-around time for that host has been. There is no guarantee that it will be consistent but if you are able to return your copy quite quickly whilst the other host shows a turn-around time large enough to put its copy well behind yours, it would be worthwhile to continue crunching yours. In this case, a good reason for returning yours is the possibility that the other copy might fail or never be returned anyway. If the other host shows a good crunching record and short turn-around times, you should just abort yours and move on to the next one.

Cheers,
Gary.

Michael Tangen
Michael Tangen
Joined: 11 Mar 16
Posts: 2
Credit: 729384
RAC: 0

Thanks much for the

Thanks much for the responses! That was quite helpful.

Best regards,
Michael

Runner • cyclist • beer and brewing geek • fascinated with space and time

Comment viewing options

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