I have several several servers running on flash drives. When I bought the cheapest bargain basement flash drives, They went belly up in a month. I have used Sandisk exclusively since then, but I presume any quality flash drive will give similar results. My current best time is 77 cpu years on a 24 thread machine, so it lasted in excess of 3 years running 24/7. My next best is 65 cpu years currently still running on a 24 thread machine 24/7. I use 16gb drives which i try to buy on sale for less than $10 US. I do not run the African Rainfall project on them as I tried on one machine and it seemed to eat up all the available space on the drive and made it read only, which essentially killed it. Both the autodock and Vina projects appear to run without any apparent problems. I run them as a live OS with no hard drive. When one fails, I simply load the same OS and replace the drive and WCG recognizes the machine as the same machine and continues on with the same stats.
Hope this helps.
Cheers

10 years ago, I started to operate 2 Phenom x6 II (6 threads) machines with SSD (Corsair at that time) for at least 6 years without any trouble.
Since I had a lot of HDD available (NAS HDD after NAS upgrade), I use those HDD instead of SSD in the crunching machines, including for the 2 old Phenom II x6.
If I would have to buy new drives, I would surely choose SSD (since 6 years, I use only Samsung SSD in my laptops).
The Raspberry Pi crunchers use a 64 GB Sandisk SD card.
Cheers,
Yves
---
PS: I always set the "Write to disk at most every" option to 600 seconds.
PPS: Between 2007 and 2009, I've tried to run the crunching machines on USB key, but at this time, it was not as reliable and performant. In the mean time, I did not make new try.

This thread's good timing. One of my Debian rigs ("Core i5-3570 (Ivy Bridge, 4C/4T) @ 3.4 GHz" in my sig) failed a few days ago. I was using an ancient Patriot Autobahn USB 2.0 32 GB flash drive...one of those tiny ones that only sticks out a few millimeters (useful in vehicles for music). It ran HOT 24/7 and that was a concern.

I got some I/O errors and Debian wouldn't boot, and the drive was in read only mode. After looking around, apparently it's a fault mode due to the internal memory controller thinking the NAND flash is done for, and the read-only mode is permanent and is only for self-preservation (so that data can be retrieved off the drive before replacement). I'm able to access the ext4 file system from a recovery environment (Parted Magic is well worth the small cost as a recovery tool!), but there's really nothing useful on it, since it's just WCG.

I generally love Samsung and only buy Samsung and Crucial SSDs, but apparently Samsung USB flash drives don't have the same quality. And I swore off Sandisk about 10 years ago after two failures (SD card and micro SD card) -- if I find those cards I'll have to recover any photos stored on them...

Since I'm a Micron/Crucial fan, their Lexar brand seems to get good reviews. I'm considering the Lexar JumpDrive M45 32GB USB 3.1 Flash Drive (LJDM45-32GABSLNA) which is $7.99 USD at the moment. Seems like no scary reviews of premature failure, and I have some older plastic Lexar drives doing fine.

I have an old laptop HDD that I might use until it fails.

FYI: I have the checkpoint period at 1800 seconds (30 minutes) on all my BOINC/WCG rigs, even the SSDs.

Hi Sgt.Joe,
the configuration you describe is indeed a "live OS" otherwise you will not have the icon for installing the OS to another storage.
Cheers,
Yves