- May 6, 2012
- 4,971
- 1,695
- 136
Hello all
I was re-purposing an older, I think first gen, SMR drive (ST1000LM035) as an external drive. Formatted it with a cluster size of 64KB. When copying large amounts of smaller files (~5MB, mostly pictures) the drive would regularly lockup for 10-20s. There was definitely some "housekeeping" going on, judging from the sound. It'd still function just fine, everything was eventually copied. Sequential writes seemed to work fine. Tested that with a large zip file.
Reformatted with the default NTFS cluster size of 4KB, everything was fine and resulted in what you'd call normal operation for a drive.
Now SMR drives employ multiple cache levels, both on and off disk. It seems that the setup is only capable of handling 4KB clusters, with 64KB clusters causing massive R/W amplification, and overwhelming the drive to the point where it regularly locks up.
Has anyone else run into this and can confirm with other SMR drives? Because if this is true, then one should only use cluster sizes that corresponds to physical sector size on SMR drives. Which f.x. would be a problem if you format an external drive with f.x. exfat, which has a default cluster size for drives of that capacity range of 64KB.
I was re-purposing an older, I think first gen, SMR drive (ST1000LM035) as an external drive. Formatted it with a cluster size of 64KB. When copying large amounts of smaller files (~5MB, mostly pictures) the drive would regularly lockup for 10-20s. There was definitely some "housekeeping" going on, judging from the sound. It'd still function just fine, everything was eventually copied. Sequential writes seemed to work fine. Tested that with a large zip file.
Reformatted with the default NTFS cluster size of 4KB, everything was fine and resulted in what you'd call normal operation for a drive.
Now SMR drives employ multiple cache levels, both on and off disk. It seems that the setup is only capable of handling 4KB clusters, with 64KB clusters causing massive R/W amplification, and overwhelming the drive to the point where it regularly locks up.
Has anyone else run into this and can confirm with other SMR drives? Because if this is true, then one should only use cluster sizes that corresponds to physical sector size on SMR drives. Which f.x. would be a problem if you format an external drive with f.x. exfat, which has a default cluster size for drives of that capacity range of 64KB.