Find Communities by: Category | Product


March 2017 Previous month Next month

Networking can limit large scale Hyper-Converged deployments, but VxRack FLEX has the answer


In my previous post, I briefly mentioned how easily the VxRack System FLEX  scales – in a couple of ways. The first is starting small, with as little as 4 nodes, and adding nodes non-disruptively as the need for capacity and compute grows. The second is how VxRack has all of the physical networking fully integrated in the architecture so it is ready to scale.


Why is networking so important in a hyper-converged conversation? Unlike traditional SANs with dedicated Fibre Channel switches, HCI relies on a shared Ethernet network to manage traffic between nodes. Therefore the network is critical to the HCI architecture and requires careful expertise and planning. When you are talking about hyper-converged at scale, if not done correctly, maintaining the same performance and availability levels becomes very complex and potential for network congestion and bottlenecks increases significantly.  Investing in a rack-scale fabric avoids the labor intensive re-engineering that can occur when environments scales.


The analysts are taking notice too.  At the Gartner Data Center Conference held in December, Joe Skorupa held a session titled “Networking Matters in a Hyper-converged World”.  A couple of recommendations at the end were:

  1. Integrate your network architects in HCIS decisions
  2. Make it clear to your vendors that broad deployment will depend upon better integration of networking.  If the network is ignored, the HCI solution could be limited.



Similar recommendations were published by the 451 Research Group.  In their Hyper-converged Market update published in December, they had an entire page within the report dedicated to the networking challenge and how it can limit scalability. This is the beauty of VxRack FLEX, it removes all of the networking challenges by integrating the network as part of the entire engineered system.

Now, let’s look at an example of what you’d have to do if you implemented a hyper-converged appliance(s) based on a specific business challenge. Changing workloads and application needs require their own unique CI solutions. While you may start with an appliance and realize the value of using appliances, you may look to reinvest the cost and resource savings into driving additional efficiencies that require you to scale your infrastructure. This is exactly what our customers told us and we came up with our unique VxRack System that provides optimal scale.


  1. You buy a small number of appliances to solve business challenge #1.
  2. Appliances arrive. You are responsible for the networking. You (a) purchase low-cost networking or (b) you use existing resources to connect everything together.
  3. The solution works fine at small scale and you solve business challenge #1. This small number of appliances are replicating to their peer appliances without issue.
  4. Business challenge #2 arrives! You now try to add the workload onto the existing system.
  5. This might work, but it might not. You are now ultimately double dipping on the original design – more workload without additional resources.
  6. Those happy replicating appliances that you had in step #3 aren’t so happy anymore. You are replicating more traffic between the nodes, increasing the network load and potentially affecting applications performance.
  7. So, you go back to your point product appliance vendor and ask for their help. The typical answer is, “You really should buy more appliances/nodes! – This will help increase performance and available resources.”
  8. Steps 4 through 7 continue – you are your point product account executive’s best friend.
  9. At some point, your network is maxed out. Either you did not buy ‘new’ with sufficient scalability or the existing infrastructure is full.
  10. In order to add extra nodes, you now try to ‘hack’ some additional networking. Yes! Maybe you try cascading switches or leverage other ‘existing resources’ to try to avoid a rebuild of the network.
  11. Eventually as you keep growing, not only are the applications on these appliances performing poorly, but it’s starting to affect the other applications not even living on the appliances. Why? Because the additional replication between all the nodes you have added over time, without any proper network-at-scale logical design, is starting to affect everything. (Insert heavy sigh…)
  12. Now the fun begins. You need to start making tactical decisions to segregate and build a new ‘bigger than ever’ network infrastructure to resolve the application performance issues.
  13. But this is a massively disruptive process, both operationally (system will have to be taken offline to re-network) and physically (need dedicated datacenter space for new racks, switching etc.).
  14. Hopefully the system you build is big enough to handle all the new incoming business requirements. But what happens if you don’t size it correctly? Make sure you hire that network expert who can. If you don’t plan correctly, you’ll likely have to re-architect the network on a regular basis depending on either (a) additional nodes or (b) additional workloads on the existing nodes


Are those 14 steps realistic? I’ll leave that to you to decide based on the needs of your environment. But if you take a look at the VxRack 1000, you’ll see why buying a rack-scale fabric up front as part of an engineered system will avoid all of those steps and re-engineering. Consisting of as many as 1,000 nodes, these racks are designed to support IT environments that are too large to be addressed by appliances, that don’t yet reach a level of scale that involves thousands of virtual machines.


Dell EMC has done all of the work with the VxRack FLEX that is pre-integrated, whether you start at four nodes and then scale to 1,000. VxRack System, with Cisco spine-leaf architecture, has the extensibility to support all current and unexpected workloads – your hyper-converged infrastructure now has a future.


If you have decided to go software defined or hyper-converged in your data center, VxRack FLEX is a great choice. Hyper-visor flexibility, integrated networking and scale capable of supporting any type of use case and application workload scenarios.  IT will now have the agility to respond to the business much faster than ever before and provide a more cohesive approach to managing all the elements that comprise a truly modern enterprise-class data center.

Follow me at @H2Ohs for the latest Dell EMC and VxRack news

Filter Blog

By date: By tag: