Hi,
I am wondering about the low-level file handling characteristics of sql server. For instance, if I have a db with a single filegroup that is 4 Gb, would it benefit performance to (prior to setting this all up, of course) format the drive with huge clusers, say 64K?
I guess what I'm wondering is--when sql server writes records to the file, I'd assume it's using the OS's file calls, so it would still benefit to use a cluster size that's equal to whatever the block size (if sql server uses 'blocks') or whatever size of chunk that sql server logically uses when it reads data pages.
Or is the difference so small or so insignificant that it's not even worth looking into this? And then would it be no difference in performance if I set up a single file for each major table, then a separate one for indexes, then maybe a few other files to hold a bunch of smaller tables? Is that too much admin hassle and/or server overhead?
--jsteph
I am wondering about the low-level file handling characteristics of sql server. For instance, if I have a db with a single filegroup that is 4 Gb, would it benefit performance to (prior to setting this all up, of course) format the drive with huge clusers, say 64K?
I guess what I'm wondering is--when sql server writes records to the file, I'd assume it's using the OS's file calls, so it would still benefit to use a cluster size that's equal to whatever the block size (if sql server uses 'blocks') or whatever size of chunk that sql server logically uses when it reads data pages.
Or is the difference so small or so insignificant that it's not even worth looking into this? And then would it be no difference in performance if I set up a single file for each major table, then a separate one for indexes, then maybe a few other files to hold a bunch of smaller tables? Is that too much admin hassle and/or server overhead?
--jsteph