Updated: Aug 3, 2021
Tales from the Field – June ‘21
"The Database Solution"
Last month we talked about “datacenter consolidation” playing the part of PowerFlex’s sixth superpower. A portion of that consolidation revolves around the customization of Oracle and/or MS SQL licensing for database workloads, especially in a two-layer architecture. Today we’re going to dive into that benefit a little deeper.
Take this customer as an example . . .
They built a dedicated virtual Oracle cluster utilizing vSphere, four PowerEdge servers and dedicated Dell SC storage. This configuration allowed for complete isolation of their Oracle cluster with the PowerEdge servers matching the core counts for their licensed Oracle Enterprise applications.
What happens when more compute resources are required? This would inevitably bump up their Oracle licensing requirements, and since this was a vSphere environment with RHEL VMs, every core of this vSphere environment would need to be licensed for their Oracle apps. At a list price of $47,500 for every 2 Intel cores, this adds up quickly.
This configuration would also be exacerbated with an HCI vSAN/VxRail or Nutanix solution. Every node in the cluster would need to be licensed for Oracle in these instances as each of these nodes would be contributing resources to the cluster. This is especially important when you consider most HCI nodes provide compute and storage resources to the cluster. If there’s a need for more storage, more compute resources will inevitably come along with that node which adds more licensing requirements for those sockets/cores, even if those compute resources aren’t needed for the application workload.
What once looked like a great isolation play has now turned into an (expensive) expansion nightmare.
The only way to mitigate Oracle or SQL licensing costs would be to deploy another HCI cluster dedicated to that other workload. And guess what happens when you build another cluster? You need to account for any overhead for storage redundancy plus add any additional nodes as required for application redundancy on that cluster. Maybe you only require 2 nodes for your database compute requirements, but unfortunately you need to start with 3 or 4 nodes to get the storage redundancy you require. This means you’re now licensing additional nodes for your database workloads that are only in place to give you storage redundancy! Instead of isolating your workloads to 2 – 2 node clusters, you’re now required to license and deploy a minimum of 6 nodes.
You’ve now created multiple siloes of administration, essentially moving away from one of the biggest benefits of software-defined solutions: consolidation.
So, the question is, could a PowerFlex solution solve this issue?
Absolutely! And the reason PowerFlex’s two-layer architecture is so popular.
First you start with a shared storage solution in a two-layer architecture. Your compute requirements would simply be layered on top of that storage in a virtual or bare metal solution. Instead of worrying about cluster minimums in an HCI config, you could dedicate compute-only nodes to your workloads in a very precise manner. If you require 32 cores for your Oracle as an example, you could deploy 2 x 16 core nodes running just vSphere or bare metal app and not worry about having to license any of the storage nodes. The S/O nodes are only running PowerFlex’s embedded OS and not Oracle binaries, so there’s no need to license them.
Storage expansions in this case are very easy and DB/application-license-friendly. You’d only need to add more S/O (storage-only) nodes to the existing PowerFlex cluster without any consideration or need for expanding your application licensing.
Likewise, any C/O (compute-only) node expansions for application resources could be pinpointed to exactly what’s required for that workload on new C/O node deployments.
The diagram below shows what that might look like in a PowerFlex rack deployment:
The purple nodes are simply C/O nodes running virtual or bare-metal Oracle application workloads. You’d license these nodes only for that specific workload.
The yellow nodes are incremental additions to your PowerFlex cluster and require no application licensing.
The overall idea is simple in a PowerFlex solution: lay down *whatever* storage is required for your initial or end-state workloads. Secondly, layer on the exact resources required for your application or DB workloads without any considerations around expansion licensing and storage redundancy requirements needed in other HCI solutions.
PowerFlex’s two-layer architecture is very powerful. It’s one of the major reasons our customers are moving to this next-gen storage solution and saving millions of dollars over other HCI solutions.