Running from RAM disk in Linux

user32094
user32094
Joined: 22 Feb 05
Posts: 22
Credit: 74648
RAC: 0
Topic 188538

I made a RAM disk with mount tmpfs and copied EAH directory there but EAH still accesses the HD every minute thus spinning it up. Is there a way to keep the HD silent while running EAH?

wijata.com
wijata.com
Joined: 11 Feb 05
Posts: 113
Credit: 25495895
RAC: 0

Running from RAM disk in Linux

Well, Maybe thare are some other processes that access HDD? (init, cron ?)
If Your machine does nothing, but only EAH, consider running some "liveCD" linux.

user32094
user32094
Joined: 22 Feb 05
Posts: 22
Credit: 74648
RAC: 0

If I don't run EAH the HD

If I don't run EAH the HD stays powered off.

wijata.com
wijata.com
Joined: 11 Feb 05
Posts: 113
Credit: 25495895
RAC: 0

Things to think, because i'm

Things to think, because i'm not sure of it.

1: EAH does some writes to disk (ramdisk), however system has to check all the path from root directory. To do it, it has to read dirs' inodes. This triggers disk read. This shouldn't trigger disk reads if there is a lot of RAM, thus inodes should be cached.
2: if You mount ramdisk on /mnt/ram, and then copy EAH to just /mnt/ram (not deeper eg. /mnt/ram/eah), then writes to file /mnt/ram/FILE triggers update of inode for /mnt/ram (which is on HDD).

user32094
user32094
Joined: 22 Feb 05
Posts: 22
Credit: 74648
RAC: 0

I have 1024 MB RAM of which

Message 8497 in response to message 8496

I have 1024 MB RAM of which 18 MB is used when I connect. It's a Debian system with 2.6.6 kernel and large memory (4 GB) support is enabled. The HD still spins up when the EAH files are in /home/user/ram/eah. Before they were in /home/user/ram.

Wurgl (speak^Wcrunching for Special: Off-Topic)
Wurgl (speak^Wc...
Joined: 11 Feb 05
Posts: 321
Credit: 140550008
RAC: 0

> If I don't run EAH the HD

Message 8498 in response to message 8495

> If I don't run EAH the HD stays powered off.

Did you try to change this parameter?

Write to disk at most every 300 seconds

You find it here.

user32094
user32094
Joined: 22 Feb 05
Posts: 22
Credit: 74648
RAC: 0

> Write to disk at most every

Message 8499 in response to message 8498

> Write to disk at most every 300 seconds

I tried this with seti and it has no effect with values like 600 or 86400.

Wurgl (speak^Wcrunching for Special: Off-Topic)
Wurgl (speak^Wc...
Joined: 11 Feb 05
Posts: 321
Credit: 140550008
RAC: 0

> > Write to disk at most

Message 8500 in response to message 8499

> > Write to disk at most every 300 seconds
>
> I tried this with seti and it has no effect with values like 600 or 86400.

Look into sched_request_einstein.phys.uwm.edu.xml in your BOINC folder. This file is only updated when you contact the server, so it might not hold the values. I have no idea if this file holds a checksum, if not you can edit the quoted parameter and restart boinc, otherwise you need to contact the server.

I once tried to check with strace what the program was actually doing, but for whatever reason this caused einstein to stop computing? Strange :-(

However. With strace -e trace=open,close,write,read,access,stat,creat,unlink -p you might be able to see the reason. But as I said, on my machine the einstein client stopped working, so you may have to restart it.

user32094
user32094
Joined: 22 Feb 05
Posts: 22
Credit: 74648
RAC: 0

strace ... -ppid reports no

Message 8501 in response to message 8500

strace ... -ppid reports no activity with boinc and prints SIGALARM and exits with einstein. strace ... boinc prints some initial activity but nothing after that but still every 30 seconds the HD spins up. I set the wait time for spin down to 20 seconds with hdparm.

Comment viewing options

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