0

Other than accidentally filling your entire C:\ drive with data, are there any other considerations for using the OS drive for data files? Most answers are simply "don't do it", but I'd like to know any technical reasons.

I'm on a dev box and will gladly trade risk for extra IO bandwidth. I'm refactoring a very large dataset, which generates 10-20 GB (simple) log files and a ton of tempdb activity. I'm moving the source, read-only tables to sata SSDs. I'd like to give the refactored data, tempdb data, and log files their own NVMe, but that means one of them would share C: with the OS.

2

For a production solution it's better to put OS, data, log, and tempdb on different volumes, even if those volumes share a single disk, storage pool, or SAN array.

This limits the blast area of running out of space, and provides separate visibility to the different IO types through the Windows Logical Disk performance counters.

However, it works fine, and is absolutely supported to put everything on C.

1
  • Ah, visibility to perf counters. I'm planning production style filegroups and files to move (or stripe) data as I figure out their IO needs. I'll look into perf counters and granularity next. – EricP Aug 28 '20 at 17:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.