Converged infrastructure is the grouping together of aspects of a technology platform in a single, optimized system. Among other benefits, this type of architecture delivers a higher level of performance. But what does this mean to an organization, and when should converged infrastructure technologies be put into action?
The first issue is around nomenclature. Storage vendors tend to use the simple term converged infrastructure (CI) to cover their products where there is a degree of compute capability alongside the storage array. Within a simple converged infrastructure platform, however, that compute capability is not there to run a general workload, such as Microsoft Exchange Server or SAP; it is there solely to run the intelligence around the array, as with what's necessary for data compression, deduplication, backups, snapshots and so on.
Other vendors have taken it to the next step. This is where hyper-converged infrastructure (HCI) systems come in. These are full systems that consist of compute, storage and networking -- all engineered together for the best possible performance.
Early entrants in this market include Scale Computing, Nutanix, SympliVity -- which Hewlett Packard Enterprise (HPE) acquired in early 2017 -- and Pivot3. And bigger players, including Dell, Lenovo, Huawei and Cisco, haven't overlooked the opportunity. Dell has its own FX2 hyper-converged boxes, and it gained VxRack via its EMC acquisition. HPE already had its hyper-converged systems before it acquired SimpliVity. Cisco has its unified computing system. Other vendors have formed partnerships so that their certified servers, storage products and networking architectures work with software from the likes of Pivot3 and SimpliVity.
But why go for converged infrastructure technologies when IT equipment is now so cheap, standardized and easily available? The key here is not that standardization makes it easier to connect different IT equipment. What matters most is the level of technical fidelity that can be achieved. Sure, a server from vendor A can be attached to a storage array from vendor B via a network from vendor C. In each stage, however, the level of standardization may require compromise.
With convergence, everything can be hyper-tuned: The platform is a known structure, and the compute, network and storage resources can be tightly engineered to get every possible ounce of performance.
This scenario appeals to businesses hoping to free themselves from complicated installation and maintenance burdens. Unpack an HCI system, plug it in, feed in a few basic details and off you go -- a fully operational system, complete with systems management to monitor and maintain the platform as it runs. For an organization that needs to respond rapidly to changing conditions, this kind of speed can make a big difference.
Most HCI systems also come optimized for cloud computing. A key to how HCI operates is in the sharing of resources across multiple workloads, so a cloudless HCI box makes little sense -- you may as well just run a scale-up large computer, such as an Oracle Sun or HPE Integrity NonStop server. With built-in resource management and workload orchestration, a good HCI platform enables an organization to move to cloud computing without needing to fully understand how to architect a cloud environment.
Capabilities and limitations
All of this considered, a converged product might sound like a perfect answer to every IT question. It isn't. Such a hyper-tuned platform tends to be specific to a workload. Attempting to run virtual desktop infrastructure, for example, alongside a transactional e-commerce application on a converged platform is unlikely to work. It's best to place similar workloads on the same system, with other systems being used for other workloads.
Some converged infrastructure technologies also suffer from a lack of single-resource scale-out capabilities. For systems that are over-engineered, running out of storage can mean having to purchase an extra system, with extra compute and network resources -- even if you don't need them. They may also require highly specific hardware from the same vendor, so watch out for vendor lock-in. Choose systems that allow for the simple addition of each resource as required.
Overall, though, converged systems make sense. Organizations would be wise to look to full hyper-converged systems, as this leads to full optimization of the total platform, as opposed to only the storage systems in converged systems.
A hyper-converged system will be especially useful for tasks such as providing centralized desktops. Running multiple desktops against a back-end architecture has taxed IT teams since Citrix first introduced the idea. HCI provides a complete setup that can all but guarantee a number of desktops that can be served. When combined with advanced desktop approaches, such as application layering or intelligent desktops, HCI can transform the provisioning of centralized desktops.
Hyper-converged infrastructure is also good for running VMs or containers -- as long as the overall workload characteristics are similar. With the built-in elastic resource and orchestration capabilities, HCI provides a solid platform to deploy and manage such pre-prepared workloads. If the workloads' needs are wildly different, however, it will be difficult or impossible for the software to effectively manage the orchestration within defined service levels.
Is there a role for straightforward converged infrastructure technologies? There is. A CI product makes storage arrays far more effective, building in lots of additional software capabilities that used to have to be layered on separately.
For an organization looking to upgrade its storage alongside existing server and network platforms, converged infrastructure will add appreciable value to the mix. Just bear in mind that storage is only one part of that triumvirate of storage, CPU and network. A hyper-converged product can add even more value to the converged story.
Knowing when it's time
Settling on the best time to move to hyper-converged or converged infrastructure can be tricky, and the factors involved in that determination will be different for each organization. Even so, there a couple of situations that may be important indicators that the time is right. One such crossroads is when a new system is required. Adding a CI or HCI product enables that platform to be up and running in the shortest possible time.
Another time to consider CI or HCI is when existing systems are up for review. Organizations should run their IT equipment against an agreed lifecycle, which ensures that problems with an old platform don't hold back the business. As such equipment comes up for review, it makes sense to consider a more integrated platform; HCI provides the best means of doing this.
Converged infrastructure technologies are becoming mainstream. Organizations should consider their use, but with their eyes open. Look for ongoing flexibility in scaling out resources, and watch out for vendors that don't seem eager to provide that flexibility.
What to expect in an HCI implementation
For IT shops, software is a factor in the CI discussion
The converged systems market evolves quickly