We're #200 @ Rosetta@Home, improving the world!

[SitFlyer]

Once my downloaded WU are done, I will be on a hiatus for a while.

Summer like temps are coming faster than spring.

[VJ]

I reinstalled the Debian client, it downloaded some work today (like last time), it should start crunching in the evening (I really need my computing power for work during the day). Last time, it just stopped crunching and refused to continue. Hopefully now it will. We'll see tomorrow...

[Umfriend]

Wulfman, you left us too?

[J1NG]

It's an entirely new SSD. Written only 170GB worth of data (applications) onto it. Never deleted anything as it was intended as a Ramdisk fast copy. (Anything I need is copied off this SSD onto the Ramdisk if I need anything with faster access or lots of constant access, eg any development tools I might need to use, gets transferred over, etc) So it was planned from the start to contain only install once and leave to copy off later aplplications and data, so wear level is virtually zero at all times since obtained.

So for all intents and purposes, 0 wear at this time. In a months time, will see how it's affected this SSD.

Also changed the write to file time on Boinc from 60s to 600s (10 minutes) to reduce writes by factor of 10 to reduce writes but keep periodic updates. 10 minutes can be made back in short order, but cycles used is a bigger issue here since my budget to replace anything has gone to zero (and will remain so for a bit for the forseeable future), so I'd rather not tempt fate.

J1NG

[Umfriend]

I can certainly understand your reluctance to run it off an SSD given the possible wear on it. But as said, I do not seem to see much of that. Of course, I am only running 8 threads on it

I am interest in your findings though. Have you taken a time=0 wear indicator (like wear levelling count or TB written)?

[J1NG]

Given up on getting Boinc working on USB Flash entirely now.

Can't get it to run more than 5 threads before it falls over itself on any USB Flash drive I have. And no longer have a HDD available since the last once has just failed as I was setting it up for Boinc.

So I'm going to stick it on the SSD and monitor usage and degredation over the next 4 weeks. Assuming it's acceptable, I'll keep it there.

9 threads and running now at 4.25Ghz each.

J1NG

[Umfriend]

Give it all you got!

[Nicram]

We will see for how long
I do not feel comfortable being 2 :P

[Umfriend]

Tazs-Matt, I think you dropped out again but in the meantime gained us 30K credits, thanks for that!

I am missing credits from Mashmallowman, VJ and J1NG. Dang the Australian summer and new build issues.

Meanwhile, I have reclaimed my #1 spot within a the team. It took me a Server rebuild and a lot of time but the natural order has been restored

[J1NG]

Was keeping an eye on things through Resource Monitor when I found the resetting of the Work Units. And found near 130 small files being access for writes each second per work unit.

USB flash drive is an older Corsair Voyager GT USB3 16GB model bought in 2012 but design and manufacture was from 2009 I think. So it's IOPS is probably not as great as it should be, hence why it's choking when I tried putting more than 5 Work Units going simulatenously. Will be trying a later confirmed model and if that fails, either find an old HDD or an ultra cheap SSD to install and crunch on. But I'm definitely not going to install Boinc onto the two existing SSD drives in this system.

Stopped getting any new work units for now, going to finish the existing ones off, then try installing onto a new storage area. Should be done by tomorrow.

J1NG

[Umfriend]

Oh, and it seems Taz-Matt has Resurrected! Don't tell him, he may not actually be aware his PCs are crunching again (and well they do indeed)! Thanks!

[Umfriend]

For what it is worth, I am running Rosetta on my Server on the SSD. I did have some concern about it wearing the SSD out but after 78 days, 2.04TB has been written in total, so that is 10TB per annum and includes install and all other OS driven writes. It is actually a lot less than I had expected and I see no reason to redirect it to a spinner anymore. And that is with 8 threads running basically full time (albeit it 3.0 / 3.1 Ghz i7-3770).

[Umfriend]

Aside from not getting credits for the team, I actually think this is funny: you are thrashing a USB Flash drive. What Class is it? And how do you know how many file operations are issued?

[J1NG]

Hmmm. Might have been a bit premature on the expunging of those last 9 units of work. I think instead I ran into a physical limitation of my existing USB Flash drive instead. With 6 work units I'm still running into the reset issue, it appears that a Corsair Voyager GT 16GB USB3 drive, is being hammered with around 900 near simultaneous file writes per second with 6 work units, and the poor drive is simply unable to cope with the demand.

The break off point appears to be approximately 750-800 near simultaneous writes (approx 130-140 separate file writes per work unit), even if it's just small changes in total (<15kb in size). Anything above this number of near simultaneous file writes and the drive will simply fail the writes and Boinc will reset the work unit to 0% again to repeat (at 9 work units, it was attempting some 1350 odd file writes in quick succession. So there is a rather bad limitation of using a USB Flash drive as the storage for Boinc. I'm going to have to see if I can't find a way to Ramdisk it, or grab an older HD to plonk it on if I want more than 5 work units at a time going.

J1NG

[J1NG]

Nah, you'll overtake easily. I'm still sorting out Boinc. Although everything is working fine, it's encountered a "constant reset" bug for all existing threads. Going to have to abort the current set of 9 and request some new ones. No new additions from me for at least a day.

J1NG