Enterprise Hyperconvergence - workload placement

In a recent discussion regarding hyperconverged infrastructures I made the following comment  "I'm partial to the hyperconverged VCE offerings for many reasons and and have suffered through the convergence decision process in several bank, telco and financial services customers. VXrail is dead simple, built on proven tech, and scales; as does VXrack. My issue with Nutanix (when it comes up) is an inability to scale and its proprietary file-system technology. Every Nutanix customer will wind up managing a lot of small clusters in the effort to go large, there are no large customer communities or developer communities behind the product or technology.  

For small installs - VXrail is a much better value, if you are going for actual enterprise grade workloads (significant Oracle DW or OLTP) a VXrack or a vBlock 540 would be a much smarter choice. I recently did a POC where we put a vBlock 540 (12c Oracle RAC and Weblogic) against an two Oracle engineered systems - the Exadata 5 (12c Exa version) and the Exalogic (running Weblogic).  The vBlock beat it soundly in terms of raw DB and application execution speed (which frankly; surprised me) - we clustered the Weblogic app servers on the same vBlock resulting in faster overall system job processing and application execution.  The vBlock was jaw-droppingly fast."

So why was it faster?  From a basic parts comparison the blades were nearly identical.  In actuality, I think it boils down to better networking technology (Cisco vs. wherever Oracle's IB tech came from) and the concepts surrounding the placement of workloads in a cloud.  Being able to cloud place the enterprise workload close to the data - both physically and virtually was a win.  I feel that our ability to mix DB and App engines on the same back-plane - on a high speed Cisco link proved out that architecture had an important place in creating a highly performant environment.  In the end we ran a 100% Oracle workload far better than Oracle at a much lower price point with increased business agility.  After all - you can put any workload on a vBlock - and this simply isn't true of Oracle's line of Engineered Systems products.  More on that later.

One important item I failed to mention was the simple ROI that comes from maintaining a uniform fleet of compute resources.  In the proof of concept mentioned above - keeping the application server and database server blades identical allowed for a friction free re-allocation of resources whenever the situation warranted.  For example, at one point in the data migration we decided to add a few more nodes to speed things up a bit.  Carving out those resources and allocating them was just a few keystrokes and resource management was a snap.  VMware vCloud Director was the key to building secure private clouds that dramatically increased efficiency and agility by pooling virtual infrastructure resources and delivering them as catalog-based services.

Fundamentally - the concepts of cloud and hyperconverged are aligned.  The best converged infrastructures will be built by strong enterprise architects who design for ROI and flexibility.  Oracle's single purpose machines, and Nutanix's promise of Enterprise Grade at Best Buy prices are both distractions.  Clearly, if you bring together the best compute, storage and networking with strong enterprise IT architecture - uniformity and cost control with high performance outcomes are easily achievable.