Number of buses on rx3600/rx6600 8 port SAS Controller?

[rhinoesterley]

If we're looking for a no single point of failure server solution, is our only option to go with two 8 port SAS Controllers and the rx6600? We plan on mirroring the boot drives, but the quickspecs state that each drive has it's own "3GB Link" but nothing about different buses. If the SAS Controller fails, even though each drives has it's own link - wouldn't they all be on the bus? Installing two 8 Ports SAS controllers on a rx6600 would solve this potential issue would it not? We're considering both server options - the rx3600 and the rx6600.

Also - can you mirror across with two 8 Port SAS drive - from disk 0 to disk 8?

Help is appreciated.

[TTr · Accepted solution]

You are on the right track when it comes to single point of failure. Yes it is true if the 8-port controller fails on the 3600 the server is down. You only mention mirroring the boot drives. If you don't need any more space and you don't need the expandability of the 6600, a better approach might be to buy two simple SCSI controllers and connect two external disks to them and mirror across the two controllers/disks. You can do two 8-port controllers on a 3600 one for the internal disks and one for an external enclosure.

You can even go a step further, Use hardware mirroring on each controller and use LVM mirroring across the two controllers on the IR-volumes.

On another thought, where do you draw the line on the SPF? If the single controller is considered a SPF, what if the motherboard fails?

So if your environment is super critical, consider a serviceguard cluster, otherwise, you can mirror across controllers but even a single 8-port RAID controller has high availability expectancy.

I would use the cost, the # of CPU's and the number of internal disks as a deciding factor between a 3600 or a 6600 and not the two internal controllers.

If you don't need the disk capacity and the CPUs, an rx2600 with dual controllers and dual boot disks would be a much cheaper alternative.