I know, before here in the forums, I noticed some people discussing the strain that BOINC and E@H can put on a harddrive, writing to disk every minute for days, weeks, years on end... I have partially solved my possibly catastrophic failures with that issue, and thought I would share it with all of you. On my desktop, I set up, a second harddrive, that I currently use primarily for BOINC, Virtual Memory and software firewall (So constant traffic, security, system and packet loging do not wear my main drive). I have my system set to turn off harddisks after 3 minutes of idle time. I currently have my 'write to disk' interval set to 3600 seconds (1 hour). But, if you restart, or shut your computer down often, it is best to set it to a lower amount of time, to avoid losing computational time... Note: I am unsure if at the same time the second drive is writing or intialized, if the primary drive is turned on to spin, or not. I will test that sometime. But, even in the case the primary drive does spin, I do not believe it is as much 'wear and tear' as if it was being written to... This is a simple way to avoid losing your OS, and other data in the event the primary disk drive fails altogether....
NOTE: It would also be possible to have BOINC run on a flash drive, or alike. Being that most USB Flash drives have enough memory to handle most BOINC projects.
d3xt3r.net
Copyright © 2024 Einstein@Home. All rights reserved.
Saving your primary harddisk
)
Dex -
Changing the duration between write cycles seems like a great idea -- I moved mine to one every 15 minutes, which is a divisor of 4 for my current cycle settings of changing projects every 60 minutes. Like you say, it minimizes the wear on the harddrive.
I'm not too sure about the use of a flashdrive though, simply because I've not experimented with it to any extent. There was a thread on this some time back, but it might have been under another project. The thread suggested that there was a finite number of times that one could reasonably expect to acces a flashdrive before a failure occurred, and that number was very low with respect to the failure point of a normal harddrive.
If I locate that thread, I'll add a link to it here ...
If I've lived this long - I gotta be that old!
For some years I've had
)
For some years I've had separate hard drives for the primary/boot and the data/applications. The impetus for this was earlier ( non-NT ) Windoze variants though, as Microsoft have only really reached adequate stability since XP in my opinion/experience. Many here will recall the lunacy/despair of an unbootable/mashed machine which needs a re-partition - possibly only due to an errant third party driver buggering up ring zero ( PlugNPray ). As you say you can flick the Windows swap file off the boot drive as well. Certainly the newer power saving features are also wear saving. I've also paid alot more attention to overall case airflow, siting of the case in the room, and separation/gap of drives in their racks. I haven't bothered with a removeable rack/cooker for years, and really won't need to now with the USB sticks. Speaking of which I suppose you could run BOINC on that, slower access though.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
You could also set up a ram
)
You could also set up a ram disk, which saves all the data automatically to the hard disk when you shut down the system. Sure you could loose crunching time, if you have a power failure. But with only 5 hours for a long WU and 0.5 hours for a short WU, this is better than kill constantly your hard disk.
I have set up a 200mb ram disk for Einstein only, with only 100mb i didn't get work with the error: not enough free disk space ..
greets
The probable cause for the
)
The probable cause for the free disk space error, was the General Preferences. Standard I believe it is set to leave 1GB or 100MB free. So, that could have been the issue. I have not set-up a ram disk since I was using a old 386 I believe. Could someone, give me a general idea on how to set-up a ram disk?
d3xt3r.net
Unfortunately not, i did set
)
Unfortunately not, i did set this to 0.0001 and use no more than ... of disk space to 100%.
I use Superspeed Ram Disk Plus.
With this programm it's very easy to set up a ram disk. Unfortunately it costs about 50$.
greets
Just out of interest what
)
Just out of interest what sort of cheap crap Hard Disks are you using?
On my dedicated crunching boxes i still have the original hard disks running. Cheap generic Maxtor and Seagate IDE33 Drives. The largest one is 4.2Gb to give you an idea of their age. I have lost 2 since i started crunching boinc a 750Mb drive and a 1.5Gb drive. Otherwise no failures yet.
I really doubt most hard disks will really feel the strain of all these disk writes for a few years yet. By that point im shure the computer they are in will be nothing more than a cruncher due to age and i would be surprised if something else hasn't died hardware wise first.
I never throw away a good piece of hardware myself so my suggestion is as those of us who have little farms retire old crunchers keep the old hard disks and other spare parts and put them into newer machines or use them to keep old machines going. I find the only time i end up buying new hardware is when i buy a new machine and often then i still use an old CDROM drive and hard disk. For example more Core 2 Duo looks like its going to get a 3.2Gb hard disk as soon as i get my next computer and the 80Gb from there will go into the new computer.
RE: Just out of interest
)
1) Another good use for old disk drives- I have a couple of small, but fast drives. I use those for swap files. Windows or Linux on the main drive, the swap on another. It seems to speed the systems up quite a bit.
2) On the suggestion of using USB drives, I would do that only if you are planning on using them as "throw-aways". I seem to recall something about flash drives have a limited number of write cycles.
Some things to keep in mind
)
Some things to keep in mind here:
The number #1 cause for modern hard drive failures is overheating, followed by mechanical damage to the magnetic media itself. This can be from causes ranging from a crash of the flying heads due to excessive vibration/rough handling, to environmental issues like particulate contamination, condensation, or corrosive gasses in the platter cavity.
Here's a pretty good quick read overview of reliability considerations for modern hard drives:
Reliability of HDD's
Note from this that in the typical home/office environment the limiting constraint is going to be start/stop cycles, which is exactly what you are talking about here.
Therefore, it is entirely possible to actually make the operating environment less conducive to maximum service life from the drive by setting parameters which inhibit or postpone drive operations and force more start/stop cycles than would otherwise occur. Of course you need to take into account what the primarly purpose of the machine is when making these choices. Obviously the operating concerns for a 24/7/365 server are considerably different, than a "road warrior's" laptop or an iPod is for example.
Based on my experience, you are far better off to take steps to make sure your drives are kept nice and cool, and stay spun up as much as possible. When you compare what HDD's draw for power compared to any modern processor or graphics card and take into account the "value" of what's on it, spinning them down can be penny wise and dollar foolish.
Alinator
Since I'm using up Harddrives
)
Since I'm using up Harddrives as "Expendables", my experiences with that technique (did that for years) :
For as long as the System is very clean and no Processes are intermittently waking up the sleeping Harddrives, this really works like a champ.
But as soon as that happens, the affected Drive can quickly suffer from a very high amount of unmonitored Power-Up cycles.
I'd test it by reading out the SMART value (Power Up cycles) off the Drive and compare it daily to the old numbers from day before.
If the System isn't touched and the number stays about the same, it works. But you note a high increase, it means the Drive is waking up frequently.
(saw that after a Drive was on its last legs and due for replacements, turned out its Power-Up cycle count was in the Millions (!) and I simply never noticed)
In my experience, such Drives will age very quickly indeed and are better left running continuously (naturally at the expense of additional Power usage, heat production and uptime).
The only Problem with Drives running 24/7 for very long times, is that they begin to "burn-in", basically 'get used' to their almost constant Temperature.
After 1-2 years, such Drives can easily be killed by simply powering them up at low room temperatures after a few days of being shutdown (been there, done that :p )
After a few days off, my typical 'Soft Failure rate' (fixable) especially on long-running 24/7 System components (even with added room heating to 'emulate' something closer to their normal temperature) is about 5%, the colder the higher the failure rate and the more likely I'm looking at a 'Hard Failure' (exchange of failed components required).
Well ... you already lost two
)
Well ... you already lost two hard disks. So i think you gave yourself the answer. In the file server that runs 24/7 sits a 200gb Samsung as primary drive, a 250gb Seagate and 3*400 gb Seagate. The hard disks shall work as long as they can, and the probability to fail increases with the work it has to do. Furthermore the hard disk can go to stand by if it has nothing to do, which saves 10 W per disk. That's about 15 Euro per year and disk.
And don't forget the work you have with a hard disk failing ... send in to manufacturer for guarantee, installing the new hard disk, restoring the image...
greets