SubnetMask
Contributor
- Joined
- Jul 27, 2017
- Messages
- 129
As I understand it (and I could be wrong), when you are using SAS disks and have dual expander backplanes, and connect both expanders to a single controller, the two connections can be combined, providing 8x6Gbps vs 4x6Gbps with a single link. In the same scenario, but with two separate controllers, this cannot happen and the second link is for redundancy only in the event of a controller failure.
In terms of reliability, with two separate controllers, if one controller were to fail, the odds of it taking down the other controller are likely low, so it would allow everything to continue functioning. I'm assuming that in the case of a single dual port controller, if something were to fail, the entire controller would likely fail, not just one link, thus taking everything down.
That being said, are there any real benefits to running a single controller vs two (Assuming all 7.2k spinning disks)? I doubt very much a real performance difference would likely be seen with spinning disks as to totally saturate the 24Gbps theoretical limit (including overhead) provided by the four links, there would need to be at least 20 of them all running to their absolute max throughput, which isn't likely.
In terms of reliability, with two separate controllers, if one controller were to fail, the odds of it taking down the other controller are likely low, so it would allow everything to continue functioning. I'm assuming that in the case of a single dual port controller, if something were to fail, the entire controller would likely fail, not just one link, thus taking everything down.
That being said, are there any real benefits to running a single controller vs two (Assuming all 7.2k spinning disks)? I doubt very much a real performance difference would likely be seen with spinning disks as to totally saturate the 24Gbps theoretical limit (including overhead) provided by the four links, there would need to be at least 20 of them all running to their absolute max throughput, which isn't likely.