I've been asked to provide my input on best practices to improve tempdb performance. I realise that this is a case by case basis and tempdb in one estate may have different requirements to another due to workload volume and type however I was hoping to get an idea of best practices beyond my current suggestions.
At the moment we have 1 data file which I will increase to 4 to match the number of logical cores. My tempdb files currently reside on the same drives as other database files, they will be moved to their own to stop IO contention. I'm also to implement trace flags 1117,1118 to grow all files equally in a file group and solely use uniform extents. Finally, a further change I'll look to make is enabling Instant File Initialisation.
Beyond what I have above, are there any best practices I've missed? Also, it suggested that storing tempdb on it's own drive is a best practice but should all tempdb files be stored on the same drive or should .ldf,.mdf & .ndf be stored on their own volumes.