On 12/23/2021 11:42 AM, Walter Underwood wrote:
When you request 5000 rows, you are requesting 5000 disk reads times the number of fields requested. It is normal for that to be a lot of disk access.
... and in order for that to add up to 3 minutes they should be around 0.03 s/read, which is only "normal" on a non-TLER drive that's about to kick the bucket.
For comparison, git does 3 disk *writes* for each file in the project, and for linux kernel w/ 60K+ files spinning rust HDDs still perform tolerably well on a directly attached TLER SATA drive w/ ext4 or xfs.
Dima
