I am currently setting up an FAS8040 in c-mode and have a big stack of SAS disks and a big stack of SATA disks and some SSD's. The architech plan that we have been looking at is once the root aggregates are built to put all SAS to controller 1 and all SATA to controller 2 with the SSD cache used where needed.
This is how we have done our previous 7-mode filers but we run against the problem of one controller is norrmally overworked and the other does next to nothing no one wants to used SATA unless they are forced to.
I am going to suggest that we split the disks accross both controllers so that they have halve the SAS and halve the SATA disks each. While this will reduce the spindle count in each aggregate we can provision fast/slow disks from both controllers to split the work load.
Can anyone give me pros/cons of laying out the disks this way or should we stick to the first layout. Our architechs plan everything to get as much space as possible without thinking about workloads etc which is why I want to look at if changing the layout might be a better plan.
Hi from a quick count up on this cluster I have 144 SAS and 72 SATA drive plus 24 SSD's this is going to be for vmware and test/dev environments , on the other cluster there is 72 SAS 72 SATA and 12 SSD's this will be for vmware and a production environment. Both clusters have a second HA pair with SATA for vaults/mirrors. We also have an exisitng 8020 on 7-mode and a plan to migrate it into the cluster in the future and plans to buy extra disks if required so these sizes will grow in future.
I agree with Eugene...it depends but often is how we setup systems. If 8.3, you can setup the SSD as a FlashPool then have 4 partitions available to both nodes without having to dedicate parity for SSD to each controller...much better usable off of the SSD this way.