We have an ~50TB volume full of faculty, staff, and department home folders. We are separating the faculty and staff home folders from the department home folders by SnapMirroring the volume, breaking it, deleting the SnapMirror snapshot, and then running powershell scripts to remove all of the faculty and staff home folders from the new home for department home folders. We have 13 servers handling two letters of the alphabet a piece, and we are seeing around 730GB/hr being deleted. Does anyone know of a way that we can speed this up any further? This is running against an A300. The new dept volume is on its own aggregate. The A300 appears to be handling this easily.
#Increase delete performance?
fringe grove
cloud oak
Doubtful as it is probably a crapton of folders and small files. For CIFS you have to have a metadata call for each IOP. If you have multiple clients spread the deletes across them
Make sure the clients are as close as possible to storage.
Make sure no firewalls or security software is running that could slow it down.
By firewalls I mean anything man-in-the-middle that modifies TCP streams, like DLP, Firewall, packet inspection hardware, etc.
ocean sleet
I seem to recall a feature…
async-delete from ONTAP
ocean sleet
I like how quick that was 😁
covert geyser
we have used it on our system in the past, with volumes in excess of 120mil files and ~40tb
fringe grove
Haven't tried it, but I will. Thanks TMAC and TyMercer!
cloud oak
Yeah the async delete is probably the best option.