tmpfs folder no faster than folder on the HD

[derekpk]

I have a large compile job running, that takes about 60 minutes to complete. I decided to create a tmpfs folder and mount it in place of the original folder.

So for example I had the following:
/home/build, I renamed this to /home/build_temp

I then ran the following commands

cd /home
mkdir build
mount -t tmpfs none build -o size=12288m

I then compiled the job again, I didn't notice any difference in the speed of the job.

Any ideas?

I'm running Virtualbox v 4.3.30 on windows 7
My VM is rehl 7.1

Thanks.

Derek.

[scottgus1]

Not a Linux guy myself, so I had to google what a "tmpfs" was. It's apparently a ram drive. Are you sure the "tmpfs" commands worked? Does "size=12288m" mean 12,288 megabytes, or 12 gigabytes? What happens in the guest stays in the guest, so to speak. If the command worked, it would try to pull from the guest's memory. Does your guest have 12 GB of spare memory?

If your guest is running, it is likely that Virtualbox is doing its job. Check with the Red Hat guys on how to get your ram drive running.

FWIW, I made a fixed-size (not dynamic) virtual drive once in a ram drive on the host made with Dataram's free ram disk software, then attached it to a guest. Response time saving to that ram-drive-based guest disk drive was lightning fast and felt faster than saving to the guest's own virtual drive on an SSD. (no I do not work for Dataram )

[mpack]

If you make the RAM disk too large, does it start paging to disk? That would be a fun result...

Even more fun would be if we had a VM log file, and so could rely less on speculation for our answers.

[derekpk]

It turns out that my process was not slowed downed by file IO but rather CPU power, or rather lack of CPU power and numbers.

To answer mpack "If you make the RAM disk too large, does it start paging to disk? That would be a fun result.."
The tmpfs never grew bigger than 8GB so page swapping was not needed as I had allocated 12GB.

Thanks for the helpful comments.