Converging the hardware and software components needed in an SMB virtualization deployment is a hot trend in the industry. Terms like “converged infrastructure”, “hyper-convergence”, “hypervisor convergence” and “software-defined (fill in the blank)” have all emerged alongside the trend and just as quickly as they were defined, most have lost their meaning from both overuse and misuse.
In this series of blog posts, we will attempt to re-establish these definitions within the framework of the Convergence Continuum below:
Before we address convergence though, let’s set the stage by describing the traditional model of creating a virtualization environment with high availability.
Build Your Own / DIY
This is typically made up of VMware or Hyper-V plus brand name servers (Dell, HP, IBM, etc.) acting as hosts and a SAN or NAS (EMC VNXe, Dell Equallogic, HP Lefthand, NetApp, etc.) networked together to provide redundancy. The DIY architecture is tried and true and effectively offers all of the benefits of single server virtualization such as partitioning, isolation, encapsulation and hardware independence along with High Availability of VMs, when architected correctly. An example architecture might look like:
The downside to this approach is that it is complex to implement and manage. Each layer in the stack adds an added management requirement (virtualization management, SAN/NAS management and Networking Management) as well as an additional vendor in the support environment, which often leads to finger pointing without strict adherence to the hardware compatibility list of each company. This complexity is a burden for those who implement a DIY environment, as it often requires specialized training in one or more of the layers involved. The IT generalist in the mid-market targeted by Scale Computing often relies on a Value Added Reseller to implement and help manage such a solution, which adds to the overall cost of implementing and maintaining.
Monolithic Storage – Single Point of Failure
The architecture above relies on multiple servers and hypervisors having the ability to share a common storage system, which makes that system a critical single point of failure for the entire infrastructure. This is commonly referred to in the industry as 3-2-1 architecture with 1 representing the single shared storage system that all servers and VM’s depend on (also called the inverted pyramid of doom). While “scale-out” storage systems have been available to distribute storage processing and redundancy across multiple independent “nodes”, the hardware cost and additional networking required for scale out storage architectures originally restricted these solutions to very selected applications.
Down the Path of Convergence
Now that we have the basics of the DIY Architecture down, we can now continue down the path of convergence to Reference Architectures and Converged Solutions, which we will define in our next post. Stay tuned for more!