I wanted to share a recent i/o performance enhancement i configured and implemented so wanted some feedback. We are definitely seeing some gains but i have few questions. FYI: our disk space is hosted via SAN using VMWARE. I basically segregated the ldf, mdf, replication(transaction relication) and tempdb into separate volume drives as below:
drive 1 - ldf + tempdb ldf + distribution ldf -> total - 10 ldf
drive 2 - mdf + replication snapshot folder + distribution mdf -> total db files: 9 mdf
drive 3 - tempdb mdf (one for each core so a total of 4) total = 4 mdf
Question
1) Is this a good approach to spreading i/o across different volumes ??
2) I have my concerns about putting replication snapshot + replication on drive 2 which is high i/o because of mdf. My thought is having a continuous replication sharing same volume as mdf does help from contention. So i am thinking should i move snapshot
folder to say drive 3? And should i move all distribution ldf and mdf on drive 3 also or leave as is?? I am open to suggestion too.
3) I am also thinking of moving non clustered index to a separate filegroup and volume. Would drive 3 be a better spot?
Looking forward to some feedback