GMP rust swap disks

[Hans Åberg]

> On 20 Mar 2021, at 20:29, Dennis Clarke <dclarke at blastwave.org> wrote:
> 
> On 3/18/21 11:31 AM, Chuck Pergiel wrote:
>> On the web page Computing billions of π digits using GMP
>> <https://gmplib.org/pi-with-gmp>
>> I found this line:
>> 
>> "One needs around 250 top-capacity rust swap disks in order to compute 1
>> quadrillion digits,"
>> 
>> I know what a 'swap disk' is, and I suspect 'rust' refers to a programming
>> language, but I do not understand the connection. Why 'rust swap disks' as
>> opposed to just 'swap disks'?
>> 
> 
> I have Seagate disks that have been spinning for over a decade and in
> ZFS mirror sets and they never fail me. I trust them far more than even
> the best Samsung SSD products. Off topic I know but I am easily annoyed
> by people that say "spinning rust".  Perhaps the SSD should be called
> something like "seized bit rot".

There seem to be no rigorous investigations, but the hard drive magnetic field is not likely to be an issue, only degrading at about 1 % a year [1], but wear and tear, and mechanical failure.

To tie it up with something that might related to GMP, a fast SSD might be used as virtual memory, maybe ten time slower than RAM, but will degrade rather quickly. This issue has been reported on the new Apple Silicon computers, which for some reasons have a lot of SSD writes. So avoid doing SSD virtual memory in GMP computations for this reason.

1. https://superuser.com/questions/284427/how-much-time-until-an-unused-hard-drive-loses-its-data