Reduce disk usage and checkpoint on Windows shutdown

Mr Anderson
Mr Anderson
Joined: 28 Oct 17
Posts: 37
Credit: 136111688
RAC: 277452
Topic 219406

Ideally these tasks should just crunch away on the CPU/GPU for hours without involving the harddrive so it could shut down or at least not be in continual use. So in an effort to reduce disk usage I increased the checkpoint time, however in Windows Task Manager I still see continual small amounts of disk activity from the Einstein/BOINC processes. My RAM usage is below 50% so I doubt that it is to do with any swap file activity. Can more data be kept in RAM to prevent this?

This led me to another issue, if the computer is restarted then I often notice that the jobs are not as far advanced as they were before the reboot, i.e. some work is lost. I assume they have reverted to the last checkpoint so if this is indeed the case, why don't the tasks checkpoint when Windows is shutting down? I know that Windows applications have a chance to save their work when Windows is shutting down although I'm not too sure about services. However I think it would be unreasonable if Windows would just stop executing them without giving services a chance to save their state.

Holmis
Joined: 4 Jan 05
Posts: 1118
Credit: 1055935564
RAC: 0

Mr Anderson wrote:This led me

Mr Anderson wrote:
This led me to another issue, if the computer is restarted then I often notice that the jobs are not as far advanced as they were before the reboot, i.e. some work is lost. I assume they have reverted to the last checkpoint so if this is indeed the case, why don't the tasks checkpoint when Windows is shutting down? I know that Windows applications have a chance to save their work when Windows is shutting down although I'm not too sure about services. However I think it would be unreasonable if Windows would just stop executing them without giving services a chance to save their state.

I'll leave the first question to one of the developers to answer but I think I can help with the second one.

A task can't checkpoint at any given time due to the nature of the calculations/developer choices when writing the app etc.
As tasks only checkpoints at suitable times in the computation loop then some work will be lost even if Windows gives the application/process a chance to checkpoint. For some types of FGRP tasks and on slower computers it can take 30 min or more between checkpoints, it's just how it is. Luckily the other apps do seem to checkpoint more often so less work is lost if it has to restart from a checkpoint.

Mr Anderson
Mr Anderson
Joined: 28 Oct 17
Posts: 37
Credit: 136111688
RAC: 277452

Holmis wrote:A task can't

Holmis wrote:

A task can't checkpoint at any given time due to the nature of the calculations/developer choices when writing the app etc.

As tasks only checkpoints at suitable times in the computation loop then some work will be lost even if Windows gives the application/process a chance to checkpoint. For some types of FGRP tasks and on slower computers it can take 30 min or more between checkpoints, it's just how it is. Luckily the other apps do seem to checkpoint more often so less work is lost if it has to restart from a checkpoint.

That sounds fair enough although it's worse than I thought. I therefore have something else for the wish list: a "Checkpoint and suspend" option or perhaps even a "Checkpoint and shutdown/restart" option. So if the user wants to restart the computer or shut it down then he/she can select "Checkpoint and suspend" and wait until all the jobs have done their checkpointing and stopped and then close Windows or have it shutdown or restart automatically.

Sometimes I start my home computer for only a few minutes for checking something and it just seems a waste for the work done during that time to be lost.

Harri Liljeroos
Harri Liljeroos
Joined: 10 Dec 05
Posts: 3610
Credit: 2902282252
RAC: 1039298

Mr Anderson wrote:Holmis

Mr Anderson wrote:
Holmis wrote:

A task can't checkpoint at any given time due to the nature of the calculations/developer choices when writing the app etc.

As tasks only checkpoints at suitable times in the computation loop then some work will be lost even if Windows gives the application/process a chance to checkpoint. For some types of FGRP tasks and on slower computers it can take 30 min or more between checkpoints, it's just how it is. Luckily the other apps do seem to checkpoint more often so less work is lost if it has to restart from a checkpoint.

That sounds fair enough although it's worse than I thought. I therefore have something else for the wish list: a "Checkpoint and suspend" option or perhaps even a "Checkpoint and shutdown/restart" option. So if the user wants to restart the computer or shut it down then he/she can select "Checkpoint and suspend" and wait until all the jobs have done their checkpointing and stopped and then close Windows or have it shutdown or restart automatically.

Sometimes I start my home computer for only a few minutes for checking something and it just seems a waste for the work done during that time to be lost.

For Windows users a Boinc Manager substitute BoincTasks has a suspend at checkpoint function. That is per task, not for all tasks, so you have to separately suspend all tasks that are ready to start beforehand. Otherwise they will start running when tasks suspend at checkpoint.

Comment viewing options

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