Virtual desktops are a very different workload from virtual servers, which should come as no surprise to anyone who has designed or deployed a virtual desktop environment. There is also a lot of variation among virtual desktop infrastructure (VDI) workloads, including different customers and different groups of users within a single customer. Some are very light loads and some are CPU intensive. Some desktops just need basic resources, while others demand far more RAM and great graphics. Choices in the way you deploy a hyperconverged infrastructure (HCI) platform will help you build the right solution for each need.
One unusual option with SimpliVity is using non-SimpliVity servers to run the desktops. These compute nodes access the storage of SimpliVity clusters to hold VDI desktops. Compute nodes expand the deployment options for VDI desktops on HCI.
In addition, SimpliVity hyperconverged infrastructure nodes all contain storage—usually solid-state drives (SSD) for performance and hard disks for capacity. Recently SimpliVity added all-flash configurations that simply have SSDs for both capacity and performance. A group of OmniStack nodes form a storage cluster and are combined into a vSphere compute cluster. The balance between storage capacity and compute capacity is important in getting the best value out of the platform. With VDI, the SimpliVity data efficiency significantly reduces the amount of physical disk space consumed. The OmniStack accelerator card in each node deduplicates and compresses VM data inline, with no negative performance impact. Desktops usually have very similar disk contents and so benefit from the built-in deduplication.
The result is that there are many VDI deployments on SimpliVity that have benefitted from adding more compute capacity. Without paying for storage capacity, adding compute nodes is a cost-effective way to run more desktops on the SimpliVity storage cluster. A configuration that includes SimpliVity hyperconverged infrastructure nodes and compute servers provides the right balance of compute and storage for cost-effective desktops.
The balance of compute and storage is not the only reason to consider compute nodes. Modern applications have become much more graphically complex, requiring hardware-based 3D graphics acceleration. To provide 3D graphics, organizations need physical servers with additional GPU cards that can be delivered to desktops. SimpliVity HCI nodes are not designed for multiple GPUs, but a few can have a single GPU card for desktops. In many VDI deployments, application performance can only be guaranteed with multiple GPUs installed inside each physical server. To meet this requirement with SimpliVity, you can choose compute nodes that support multiple GPUs. NVIDIA’s hardware compatibility list includes servers that can hold five additional GPU cards, for maximum desktop graphics performance.
Compute nodes add a lot of flexibility to VDI on SimpliVity. More compute can be added to maximize the utilization of the storage and support more desktops, and highly graphical desktops can be accommodated through multiple GPUs in compute nodes. The flexibility to support very different VDI workloads is important. SimpliVity compute nodes add a dimension of flexibility that is rare in HCI.