Nearly everyone uses hyperconverged infrastructure, or plans to. A survey by InteropITX found that only 20 percent of survey respondents had no plans for using hyperconverged infrastructure. The ease of support and scalability are benefits businesses can't overlook.
Hyperconverged systems gain that ease and scalability from the integration of compute and storage into a single commodity unit. But this integration means that compute and storage are linked together—they can't be scaled separately, which eats away at some of the expected cost savings. Now some hyperconverged infrastructure vendors and customers are looking at a disaggregated model that offers the benefits of independent scaling without losing the ease of support.
Resource Pools for Composable Servers
Rather than a single unit providing both compute and storage, disaggregation pools the computers' resources and allocates memory, storage, CPU, and other resources on demand to meet application needs. The resources are neither siloed nor tied to single servers, offering even greater flexibility. Utilization levels can surpass 80 percent, while automation decreases management costs and enables infrastructure to be redeployed extremely rapidly.
Through disaggregation, servers can be treated as composable—assigned exactly the resources the application it runs needs. Unused resources are no longer locked within a specific server but are available to all servers that need to expand their capacity. This means spare capacity can now be measured and managed across the data center, rather than on a server-by-server basis.
Hyperconverged, Disaggregated, or Both Depending on Use Case
Choosing the hyperconverged or disaggregated architecture as your solution depends on the use case you're trying to solve. For most mainstream purposes, hyperconverged infrastructure continues to offer real benefits. Disaggregated makes sense for specific scenarios where the jobs are intensely computational and moving storage processes off the server provides measurable benefits. Disaggregated hyperconvergence is similar to private cloud but retains the standardized appliance model, even if the appliance isn't always used for all servers.
In fact, although superficially hyperconvergence and disaggregation seem to be totally opposite approaches, if you look below the surface, disaggregation actually extends the idea of hyperconvergence to make environments even more scalable and dynamic through adding simple components. Disaggregated hyperconvergence makes it even easier for enterprises to develop true web-scale environments. The separation of resources doesn't result in silos that make large data centers difficult to manage, unlike traditional infrastructure.
Bringing Disaggregation Into Your Data Center
Hyperconverged infrastructure vendors offer various solutions to allow you to introduce disaggregation into your data center. Nutanix nodes can be designated as compute only, including no disk storage. Nodes without compute at all aren't possible in Nutanix, but nodes with very low compute and high storage can serve that purpose. Since clusters don't need to be built with identical nodes, it's possible to craft and extend clusters with the exact level of compute and storage resources needed.
Is hyperconverged or disaggregated infrastructure more suitable for your data center? If you need some help considering the options, contact dcVAST. Our experience with hyperconverged infrastructure including Nutanix and Cohesity, plus our cloud and storage expertise, means we have the background and knowledge to give you true insight into the architecture that will help your business reach its goals.