In my previous post, I justified why to create the vSphere Management Cluster. In this short post we discuss vSphere Clusters: Scale Up and Scale Out. What's difference?
Scale Up - vSphere cluster based on larger ESXi hosts (adding more processors and RAM)
Scale Out - vSphere cluster based on smaller ESXi hosts (servers with less processors and RAM)
| Type of cluster | Pros | Cons |
|---|---|---|
| Scale up | - Costs! (licensing, power consumption) - Better utilization Transparent Page Sharing (TPS) - Ability to support larger VMs - Less chance for CPU contention and better memory performance (larger NUMA node size) | - Host failure causes, a larger number of VMs are impacted because they needed to be restarted - Higher HA overhead - Potential for Network or I/O problems due to larger number of VMs per host - Less hosts = less flexibility for DRS to load balance VMs between them |
| Scale out | - Host failure does not cause, a larger number of VMs are impacted - Less HA overhead - More hosts = greater flexibility for DRS to load balance VMs between them - Easy upgrade ESXi hosts (firmware, vSphere etc) - less VMs to evacuate | - Costs (additional servers, licensing, power consumption, more network ports required) - Less utilization of TPS - Possible CPU contention (less cores) |
So which one to use? It depends on for example: configuration of VMs, type of servers (blade or rack) or security standards (e.g. DMZ servers).
I prefer Scale Out Clusters based on Blade Servers. Easy upgrade, "be green" 🙂 and a smaller number of VMs are impacted and have to be restarted when a host fails.