Why Cisco HyperFlex is a Big Deal

By Kent Christensen
3/1/2016

Typically when I write a blog post it is more about observations from the industry, clients and events and less about a specific product (because in general product companies have people to do that). However today Cisco announced a new solution that is not only noteworthy but will be very impactful. Today Cisco announced a Hyperconverged solution called HyperFlex Systems that not only addresses the needs of the hyperconverged market, but integrates into an infrastructure that many clients are already leveraging.

Hyperconverged infrastructure has been getting a lot of notice not just because it is new, but because it addresses a specific business outcome in the IT space. Cisco has built a dominating presence in being at the heart of most of the leading Converged Infrastructure that many organizations from large to small entrust to build out a datacenter strategy that scales, reduces cost, and reduces risk for the most mission critical applications. However there is a growing number of use cases where infrastructure for lighter workloads needs even more flexibility to grow incrementally from small to very large yet simple enough to run in small data centers without a large IT staff. For example:

  • Remote offices with limited staff.
  • VDI or VSI which in aggregate support a large number of workloads that have commonality amongst them.
  • Less mission critical applications that don’t require the high performance an all flash converged infrastructure provides.

With today’s announcement Cisco provides a hyperconverged solution that compliments the portfolio of converged solutions and is much more complete than you would expect from a “first generation” product offering.

With a hyperconverged solution there are several tenets we would desire for a complete system:

  • Simplicity – It needs to be simple to implement and operate. HyperFlex can be set up in minutes and managed from common interfaces like vCenter.
  • Availability – The workloads are still mission critical and it must be resilient and highly available. HyperFlex provides real-time redundancy that is configurable to your tolerance and can withstand any component (disk) or even complete node failures.
  • Flexibility – We want to be able to tune the infrastructure between high capacity, high compute and high memory workloads. This is accomplished with offerings that address high compute, high capacity and even adding compute only nodes.
  • Efficiency – It is common to want deduplication and compression to take advantage of the redundancy seen in the targeted workloads. HyperFlex provides very efficient inline dedupe and compression that is always on.
  • Scale – We want to scale from small three – four node systems to very large clusters without compromise.

This last point warrants some specifics of why HyperFlex is different than many existing solution. The “special sauce” is a purposely built log structure scale out file system that was developed by SpringPath. The reason that is important is:

  1. This allows the system to do inline dedupe and compression so efficiently that it is always on with no impact to the application and thus is providing very high performance, lower cost and higher scalability.
  2. The log structure file system then allows large numbers of clones with minimal impact on capacity and performance. Which is what you want do with a solution like VDI.

Finally the network was not an afterthought and instead integrates to your existing network via Fabric Interconnect similar to other UCS environments. We have seen many “scale out” solutions hit walls as eventually there becomes a node to node networking problem. By integrating 10GB into each node with traffic shaping inherent in a unified network HyperFlex has even eliminated the need to require data locality to provide performance for applications. Instead as nodes are added HyperFlex automatically spreads the data evenly across all the available capacity since accessing any node in the network has no impact on the application.

This is an impressive start for a new architecture that complements Cisco’s Data Center strategy. We expect scale to grow as validation is extended (currently 3-8 nodes). Additional integration with cloud components like UCS Director also make HyperFlex an ideal infrastructure for Datalink’s Cloud Complete strategy for secure hybrid cloud.

By Kent Christensen – Practice Director Virtual Datacenter and Cloud – Datalink