High workloads on HCI

High workloads on HCI


The vision for HCI is to make the datacentre easier. Each node within a HCI cluster has everything that a data center will need as it expands; compute, hypervisor, storage software, performance storage, capacity storage, and networking. Initially, with the first three nodes, all is well. However, as the data center grows and adds nodes, the HCI vision begins to get foggy.

Most HCI vendors know that nodes are limited in the number of workloads they can handle and only recommend running certain workloads per node.

OLTP on Software-defined storage

When running OLTP workloads on an HCI infrastructure make sure you have the storage like a traditional block SAN running at the backend or if you do not have a SAN in your HCI solution then make sure you build your HCI with NVMe backends so you can get the storage to handle the OLTP transactions. I have always been a skeptic with the likes of Nutanix and those guys when it comes to high-performance applications as they have a normal software-defined storage solution.

There is a new sub-category of hyper-converged storage that has come into the market (2019), what IDC calls disaggregated hyper-converged, where the storage is on a separate appliance from NetApp, Dell EMC, IBM, or Pure Storage. By breaking storage free from the nodes, compute and storage can be scaled independently from each other – one of the big drawbacks of earlier hyper-converged systems. I have also read articles where pure SDS solutions backed out of testing their stuff on high workloads. I can at least give them a thumbs up for their honesty. This is according to me a great advantage for hyper-converged as you can now have an old-school SAN (much better today) behind the scenes doing the hard work.

The storage supporting virtual infrastructures are getting pushed harder than ever as OLTP workloads are virtualized. HCI vendors typically recommend the customer creates a cluster for each of these workloads, or at a minimum, dedicate a node. This makes supporting mixed workloads and scaling beyond initial configuration more complex.

What to look out for

Always check out if your HCI solution has iSCSI for storage and if so, you can bet you will have issues getting proper throughput for high OLTP systems. Most of the time anyway. Be cautious when your storage and network traffic runs over the same ethernet network and also with the inter-node communication between your HCI cluster nodes, more nodes do not always make things faster as communication needs to happen between nodes and there are latency as well! For general purpose stuff and most workloads, the above would be enough.

You also get some HCI vendors like Virtustream build specifically for high-end apps, but it is a Dell EMC/VCE Vxblock at the backend with an added-on appliance to do the billing. Virtustream has got a complicated thing going as their support is reliant on VCE and then Virtustream support, lots of support layers behind the scenes… It is always a headache to see executives make decisions on the vendors selling them a solution and the solution never works as described.

To me, it always depends on the workloads of your business before deciding on HCI.

Always remember these vendors spend time and money on their sales, while they have some good solutions out there you will have to equip yourself and get some proper technical people involved while making decisions to buy HCI.

By Cobus Smit – 11 Set 2019

Spread the HCI Information
  • Post Author:
  • Post published:September 11, 2019
  • Post Category:Home