This just absolutely blew my mind so I figured I’d post a quick note about it. I have to convert a CVS repository to SVN and this is a big repository, over 33k commits to it spanning a decade. In my trial runs, using cvs2svn resulted in it taking just shy of 11.5hrs to complete, on the HDD (I have an SSD in my workstation for everything and a 2TB HDD for storage and scratch space which is where I was running the conversion). Moving the original raw CVS repository into a temporary directory on the SSD and running cvs2svn from there (reading from and writing to the SSD) resulted in 19 minutes.
That’s right… 19 minutes. Moving from the HDD to an SSD dropped the conversion from 11.5hrs to 19 minutes. I expected a speed-up, but I did not expect it to be this drastic. It makes me wonder how much faster having my Fedora mirror on an SSD would be when doing rq imports. There I have the MySQL database on the SSD, but all the rpms are being read/unpacked on the large HDD. Importing can be slow, given the size of Fedora (queries to the database is really fast).
Unfortunately, my Fedora/CentOS mirror is almost 800GB and that would be a very expensive purchase. This makes me sad.