What is a Hyper-converged Infrastructure
HYPER-CONVERGED INFRASTRUCTURE (HCI) - Why all the Hype?
by Terrence A. Weadock
In the beginning, there were network servers. Each server would run its own operating system, have its own local storage, and would run dedicated application software. Various fault-tolerant features including redundant hard drives, redundant power supplies, and error-correcting (ECC) memory kept these servers running in the event of the most common component failures. Data backups were key as the question was not whether the servers would fail, but when they would fail. When they did fail, multiple day down-times were all too common.
Using consolidated storage in a “Converged Infrastructure” came next. In a typical converged infrastructure, a “storage area network” (SAN) would be used to store all the data for all applications to be run on the network. All server apps are then virtualized using Hyper-V or VMware and allocated to specific “Compute” servers. These servers would be light on disk space, but heavy on CPU and memory capacity. If a compute server became unavailable for any reason, the virtual machines could easily be reallocated to a different compute server either manually or automatically. This concept offered better fault-tolerance and more rapid recoveries but if the SAN itself went down, the whole network went down. In addition, because apps are allocated to specific compute servers, the load could not be spread evenly across the servers. For example, compute server A may have a 90% utilization on average, while server B may have a 50% utilization on average. Generally, it was difficult to efficiently allocate resources using this model.
With the newest version of Microsoft Windows Server 2016 Data Center Edition, a hyper-converged infrastructure (HCI) can be built. The easiest way to understand how HCI works is to consider that it is similar to a RAID 5 or RAID 6 disk subsystem except that instead of using redundant disks, it uses redundant servers. In this way, an HCI with 3 server nodes could have failures on 2 of its nodes and still function without a loss of data or computing horsepower. In addition, the system acts as one large server and allocates on the fly which actual node runs specific software chunks at any given time. Gone is the situation where one app server is being hammered by a demanding app like SQL Server, and another one sits idle much of the time because it is just managing file and print requests.
Over the next few years, organizations that cannot move all their systems to the cloud will find themselves wanting HCI for the following reasons:
1. Better resiliency and faster system recovery
2. Better allocation of system resources
3. More cost efficient than a converged infrastructure
Click here for a comprehensive set of hyper-convergence related links with additional details on how to specify, configure, and install an HCI system.
While this is a new technology, Dominant Systems is committed to selling and supporting HCI using HPE systems and Microsoft Windows Server software. Please call 734-971-1210 or email us at for more information on how we can bring the benefits of HCI to your organization.