Increasingly, organizations are looking to hyper-converged infrastructure (HCI) to simplify their on-premises architectures to make them more cloud-like. Obvious considerations for IT professionals looking to move to HCI are Cisco HyperFlex, Dell EMC, Nutanix, and VMware HCI. This writeup is focused on VMware HCI, and vSAN which is included with its hypervisor. The challenge with vSAN is in understanding how current workloads will perform in the HCI context. Another challenge, after conversion to HCI, is in knowing when a performance problem is on the horizon. The difficulty stems from the high level of abstraction common with HCI.
HCI can make IT professionals quickly long for the days of bare metal systems, where identifying the source of a performance problem was relatively linear. VMware, by itself, makes problem identification and resolution more difficult because it enables multiple virtual machines and hypervisor hosts to access the same storage and same LUN/Volume at the same time. In the typical VMware context, though, the storage still stands independent from the VMs and hosts. However, in the HCI context, storage loses that independence.
Within the HCI context, the physical hosts now perform everything from providing computing power to the VMs, to capacity to the storage tier, and performance to the storage software layer. Determining where a performance problem stems from is very difficult, and forecasting the impact of moving additional workloads into the HCI framework is almost impossible.
As a result of the lack of visibility, most organizations throw additional hardware at the problem by adding high-performance nodes with high-performance flash, hoping that performance problems never occur. The reality is, though, that performance challenges eventually always appear, and IT has no early warning system to address the issue before it impacts application performance.
Turbonomic 6.4 – vSAN Visibility for Enterprise-Class HCI
Turbonomic, who we covered in a previous briefing note is an application resource management solution that provides end-to-end visibility using a unique supply and demand economy model to decide how it allocates resources. They also offer a single platform for on-premises clouds and public clouds like Microsoft Azure and AWS.
The latest release of Turbonomic provides a wealth of new features, but one that might be of the most significant interest to Storage Switzerland readers is the release’s support and integration with vSAN into the software’s “economic model”.
Release 6.4’s integration with vSAN is designed to make sure the organizations are getting maximum performance and value out of their vSAN investment. Not only does the solution monitor and warn of performance problems, but it also can, with permission, take corrective action or advise IT when it is time to buy more nodes/storage. The 6.4 release is fully vSAN feature aware, which means the application resource management capabilities understand vSAN performance, capacity redundancy, compression, and deduplication functions when it makes decisions.
Turbonomic can decide which applications, VMs, or containers to place on vSAN as well as how and when to scale the vSAN infrastructure to increase IO performance, reduce latency, or meet capacity demands. The analysis is done in real-time and provides IT administrators with plenty of advanced notice when new hardware is required.
IT planners can also use Turbonomic for planning and modeling the impact of moving additional workloads to vSAN. Since Turbonomic monitors the entire environment, it is capturing actual, not simulated, data and then incorporating that data into the vSAN model, which of course it is also monitoring. The result is a very accurate prediction of the impact of moving the new workload to the HCI environment to prevent application performance problems. Turbonomic can also advise on additional hardware considerations to help compensate for the new workload.
StorageSwiss Take – Helping HCI Keep its Promises
vSAN, in particular, and HCI (including Cisco HyperFlex), in general, makes a significant promise to IT: that their infrastructure becomes more cloud-like when the organization shifts to it. Certainly, HCI automatic scaling by adding nodes does lay the foundation for a more cloud-like experience, but only if the customer is willing to accept massive over-provisioning to compensate for the lack of visibility into how HCI is using the various resources available to it. Over-provisioning is the antithesis of creating a cloud-like infrastructure.
Turbonomic eliminates the temptation to over-provision by providing real-time analytics on resource consumption, even in the heavily abstracted HCI environment. The solution does more than provide telemetry data. It can provide recommendations to prevent potential problems from occurring. As the organization builds trust in the solution, IT can enable Turbonomic to execute its recommendations automatically. Turbonomic also empowers the organization to use the same supply and demand based application resource management software in public clouds like Microsoft Azure and AWS. With the combined solutions, the organization can manage its hybrid-cloud under the same economic principles.