Separating Compute And Storage In Hyperconverged Infrastructure

Back in its earliest days, hyperconverged infrastructure was seen primarily as a consolidation play, a way to bring together compute, storage, networking and management together into a single package and offset some of the rising costs and complexities in enterprise datacenters. And unlike converged infrastructure offerings, HCI pools compute and storage into a single unit within an appliance sold by hardware established OEMs like Dell, Hewlett Packard Enterprise, and Cisco Systems as well as smaller pure-plays like Nutanix.

The role of HCI has evolved as the market has matured, with hyperconverged infrastructure becoming a key asset in the rapidly growing hybrid cloud scene as enterprise demand has grown for more cloud-like capabilities, such as automation and operational simplicity. The software that goes into these systems is now as important as the hardware components, with companies like VMware and Nutanix – which has shifted to becoming much more of a software vendor with a greater focus on OEM partnerships and subscription sales – playing a much more central role in HCI.

All this has contributed to an HCI market that continues to grow, with global revenue jumping 7.4 percent year-over-year in the fourth quarter 2020 to $2.5 billion and the segment accounting for 54.2 percent of the larger total converged infrastructure (CI) market, which also includes integrated infrastructure as well as certified and reference systems, according to IDC.

At the top of that list is Dell Technologies, which with VMware engineers the VxRail systems that leverage Dell PowerEdge servers as well as storage and HCI System Software. It also includes VMware’s virtualization technology and VMware Cloud Foundation (VCF) hybrid cloud platform as well as NSX-T networking and VMware’s Tanzu Kubernetes software, another nod at the evolution of the role of HCI in organizations’ IT environments. Dell owns 32.8 percent of the market and in the fourth quarter saw revenue jump 11.1 percent, to almost $802 million, IDC said. HPE is a distant second, with 13.5 percent market share.

“While consolidation certainly is still a driving function, really it’s about operational simplicity and driving as much automation and putting more of the work on the vendor to deliver IT that is much more simple to maintain and operate,” Shannon Champion, vice president of product marketing for primary storage for HCI and CI at Dell, said during a recent press briefing, adding that “customers tell us they really are valuing the operational simplicity and wanting to move their IT value up the value chain. In terms of a hyperconverged driver, it’s more so about that than it is consolidation in and of itself.”

Dell, which has more than 12,400 VxRail customers, this week is running out with new capabilities and features for VxRail as it looks to press its advantage in the market and highlight its deep collaboration and engineering efforts with VMware even as it moves to spin out the company later this year (Dell owns about 81 percent of VMware stock).

Included in the enhancements are the expected upgrades to the hardware and software, but at the same time the company is offering VxRail dynamic nodes, changing what has been a staple of HCI appliances – the merging of the compute and storage – by enabling enterprises to separate the two within VxRail to make the systems run more efficiently both on premises and in hybrid cloud environments.

“Depending upon a customer’s environment and the workloads that they’re running, some workloads could be more storage-heavy and some of them can be more compute-heavy,” Rick Reddy, senior director of product management for HCI and CI at Dell. “With hyperconverged, in the past you had to scale the nodes out in a relatively even fashion such that when you had a compute-intensive workload, if you needed more compute power … you had to add an additional node that included storage, although you may not need that. The same was the reverse – if you had something that was very storage-intensive and you were running out of storage, you had to add an additional node that had compute in it, but you may not need those additional compute resources. What this allows us to do is to separate the two and even though we’re separating the two into nodes that allow you to scale compute independently and nodes that have the storage and the ability to add more storage to them, they’re still running the same VxRail system software and they have the whole lifecycle management and the validation.”

VxRail appliances can now include external storage options by leveraging VMware’s vSAN HCI Mesh, which shares VMware vSAN sorage capacity across clusters. If there is unused vSAN capacity in one cluster, that now can be used by another cluster, Hurley said. At the same time, VMware Cloud Foundation on VxRail can also take advantage of dynamic nodes to use Dell EMC storage products at primary storage, including PowerStore, PowerMax and Unity XT.

“If there’s any [unused vSAN] resources within any existing VxRail cluster, what we can do is use these dynamic nodes to a create compute dynamic cluster that can point to that unused vSAN storage and that’s more effective use of your overall environment,” said Nancy Hurley, senior manager of product marketing at Dell for HCI and CI. “On the other hand, a number of customers may want to address some data-centric workloads in a VCF on VxRail environment. There may be some workloads that in the past they may have wanted to keep on [a storage tier] or they may want to use some of the data services that are available in some storage arrays, such as PowerStore and PowerMax and Unity. What this enables customers to do is to have the extra storage attached as primary storage and still utilize the operational model that you can get with VCF on VxRail.”

On the hardware side, Dell is moving VxRail to the latest PowerEdge systems that include Intel’s “Ice Lake” Xeon SP processors and optional Optane 200 Series persistent memory – delivering 32 percent more bandwidth than the previous generation – and AMD’s “Milan” Epyc 7003 server chips. In addition, the VxRail V Series appliances offer Nvidia’s A40 or A100 Tensor Core GPU accelerators that, combined with the Nvidia AI Enterprise software suite and NVM-Express protocol, are aimed at advanced artificial intelligence (AI) and machine learning applications.

The P Series systems have 20 percent more capacity for such workloads as virtual desktop infrastructure (VDI) – which has seen demand increase in the past year due to the sudden shift to remote work during the COVID-19 pandemic – and E Series have 50 percent more PCIe slots to accommodate such as network or Fibre Channel cards.

Software updates include tools in the HCI System Software – including Configuration Portal and Node Image Management – that enable organizations to self-deploy clusters on their own schedules, including validation and orchestration, which is important at a time when enterprises are managing highly remote and distributed environments. At the same time, Dell is moving its lifecycle management capabilities up the stack to include NSX-T and Tanzu in a single upgrade cycle and upgrading it with an Nvidia AI Enterprise and VMware installation bundle.

Hurley said VxRail users increasingly are adopting Nvidia GPUs, adding that “first, we’d see this just for perhaps VDI implementations, but now we’re seeing a great deal of adoption for AI and ML and a number of other applications. We’re really focused in on expanding that ease of use in the lifecycle management to the Nvida GPU. Now customers can get the bit from Nvidia for whichever of the GPUs they are utilizing, upload that into our lifecycle management and easily update their GPUs as well.”