Research Report: The Virtual Public Sector

HyperConvergence: The Next Wave of Convergence

When agencies need to spin up a new application, start a new project or simplify the infrastructure at a branch office, some are turning to a converged infrastructure instead of traditional data center resources. Doing so can simplify management, improve efficiency and lower costs.

There are times, however, when even a converged infrastructure isn’t enough—not enough performance, automation or integration. In those cases, there is another, related option: the hyper-converged infrastructure. While converged infrastructure provides integrated compute, storage and networking components designed to work together, the components often come from different vendors. Converged infrastructure also doesn’t include a hypervisor or software-defined storage. A hyper-converged infrastructure provides compute, networking and storage from one vendor, full pre-tested and configured for the intended workload. It includes the hypervisor, and uses software to eliminate the need for external storage.

Inclusion of the hypervisor is critical, because it sits between the underlying infrastructure and the applications. That means that it can abstract all underlying resources—a major key to efficiency.

Hyper-converged infrastructures are built to specifications and shipped as a single unit. For example, an agency that needs to spin up 500 Microsoft Exchange users would be able to choose a pre-configured, pre-tested hyper-converged infrastructure designed to do exactly that. All they would have to do is plug it, and it would be ready for use. If more users needed Exchange, an agency could simply buy another module for the system.

The hyper-converged infrastructure provides excellent performance and convenience for the price. It can track which data belongs to which virtual machine, making it much easier to move virtual machines around. It also eliminates redundant read and write operations; routinely includes deduplication, compression and data optimization; and takes automation to a new level. This further reduces labor-intensive activities and prevents over-provisioning and over-purchasing.

There are other benefits to the hyper-converged infrastructure. With less physical hardware in the data center, power and cooling costs are reduced, and IT staffs don’t have to spend as much time on maintenance, patches and troubleshooting. Performance tends to be higher because the components are so tightly integrated and tested. Because the storage is fully integrated, for example, there tends to be less latency. Since the networking also is integrated, there are fewer bottlenecks and better throughput. It’s also much easier to scale the infrastructure because of the building-block approach. And because the entire infrastructure comes from one vendor, there are fewer vendor relationships to maintain.

One example of a hyper-converged infrastructure is VMware’s new EVO line, consisting of EVO:RAIL for branch offices, VDI deployments or workloads like test and development; and EVO:RACK for large deployments. Both are preconfigured, pre-integrated infrastructure stacks available from vendors like Dell, Hitachi Data Systems, HP and NetApp. Both EVO:RAIL and EVO:RACK use vSphere, along with vSAN and local-attached storage.

The EVO system also results in higher OpEx and deployment savings, due to a consolidated 2U/4N platform, rapid deployment, delivered patches and automated patching and upgrades. To prove the point, VMware conducted an analysis to confirm the cost savings of this approach by comparing the building of a vSphere cluster with vSAN and vCenter with four servers from bare metal servers with the EVO approach. Based on a $250-per-hour rate for a senior consultant, the traditional vSphere cluster took 200 hours at a cost of $50,000, compared to the same with EVO:RAIL. The EVO:RAIL approach took seven hours at a cost of $1,750.