I'm configuring a new 3-node VMware cluster with R760s connected through Fibre Channel. The ME5024 has 20 spinning 2.4TB HDDs and 4 1.6TB SSDs. I'm considering creating one large pool on Controller A with ADAPT for the HDDs and RAID 10 for the SSDs. This would simplify management since I could maintain a single datastore. However, I'm wondering if having Controller B remain idle (essentially operating in an Active/Passive mode) will noticeably impact performance, given that I only have 20 HDDs. Alternatively, should I follow the 50/50 split recommended by the configuration wizard? I didn't have the chance to purchase extra HDDs, so I'm somewhat limited. My plan is to dedicate one cluster for regular VMs, including SQL databases and applications, with Controller A handling the SSD RAID 10 and ADAPT settings, while letting Controller B sit idle. For the other cluster focused on Cisco ISE, I'm thinking of a setup with RAID 10 on the SSDs and a combination of spinning disks in the second controller. Any thoughts?
4 Answers
How much data do you plan to store on the SSDs? Regardless of using one or two pools, the cache syncs between both controllers, so you'd be taxing the memory equally either way. The real advantage of having separate pools might be redundancy; if something goes wrong with the cable or SFP, only one pool would be impacted. Just remember that SSDs can fill up quickly, and when they do, performance can dip if you're relying on spinning disks.
That setup sounds reasonable! Just ensure your data is partitioned well between SSDs and HDDs. Based on your drives’ specs, with RAID 10 for SSDs and ADAPT for HDDs, you should be good. Just keep an eye on Controller B’s idleness; some folks might think it’s an underutilization, though.
In my experience with SQL DB clusters, it’s beneficial to use two separate pools. We have a similar setup over two clusters, and Dell recommended this to us to optimize performance. Your situation might be different, but it worked well for us.
With just 20 disks, it might be better to go for balanced pools across both controllers. Leaving one controller idle can lead to a bottleneck, especially under higher loads.
Got it! So you ended up using one single pool on each of your clusters, right?