Advanced Nutanix: Cluster Expansion aka “Add Node”
In the post I’ll explain how the “scale out” capabilities of the Nutanix cluster work and the process for cluster expansion (aka “Add Node”).
From my perspective one of the most important capabilities of the Nutanix cluster is it’s scale out capabilities or seamlessly scale in a granular manner. Here’s a video of me demonstrating the simplicity of scaling out a Nutanix environment:
True Convergence Enables True Scale Out
Prior to joining Nutanix I did a great deal of large data center and service provider deployments using VCE’s Vblock and NetApp’s FlexPod which are both modular pod based deployment mechanisms (obviously there are some substantial differences between them, it’s assumed the reader is aware of these). At the time the simplicity and capabilities was unlike anything we’d ever seen before and set the baseline for modular scale. However, the issue with these was that they were large pools of siloed resources with an abstracted provisioning and management layer and faced the same issues as traditional platforms.
The issues with this model were a few fold:
- It required a great deal of CapEx even for the smallest deployment
- The unit of scale was extremely large meaning even the demand for 1 more VM could mean having to scale a pod providing 000’s of VMs
- Pods were silos and resources weren’t truly shared
Also, let’s not forget the scaling of the management aspects including the management domains, UIs, logging, etc.
Here we show what an example pod looks like from a traditional vendor which can range anywhere from half a rack (20U+) to multiple racks (120U+). The pod based model takes standard components each of which has its own resource manager (eg. UCS:UCSM, Cisco Nexus:DCNM, Netapp:OnCommand, etc.) and in some cases which includes a “pod level” or abstracted resource manager (UIM, etc.).
With the Nutanix model storage, compute and management is all converged into a single node using the network as a simple fabric for communication.
Size does matter…
Actually, it does on multiple levels. No… I’m not referring to what you’re thinking about but in terms of scale. It brings me back in the day to basic principles of economics and supply chain where you want to efficiently match supply to demand as much as possible to drive higher efficiency – the same holds true in terms of IT and cloud.
The smaller scale we can start with means the smaller the upfront CapEx we require. This means you can “start small and scale” as compared to the other model of “start large and wait”. With larger step increases there can be a great deal of time where supply greatly outnumbers the actual demand. Sometimes this can be OK in very volatile or rapidly growing environments, however in most cases it just leads to “sunk” resources and inefficient utilization of capital.
With the Nutanix model of start small and scale you can minimize your CapEx and push off some of the scale to OpEx making the approval process much easier. Also, since the steps are much more granular you can more accurately match your actual demand leading to consolidation, higher utilization and higher efficiency.
In another post I’ll talk about a few of the technical items behind how the Nutanix platform can scale out so effectively.