There is no shortage of options for enterprises when it comes to Kubernetes platforms. Red Hat has OpenShift, VMware has Tanzu, and SUSE has Rancher. And the cloud providers have their own Kubernetes services, too. Specifically, Amazon Web Services has Elastic Kubernetes Services (EKS), Microsoft has Azure Kubernetes Services (AKS), and the Google has Kubernetes Engine (GKE).
That diversity should not be surprising. Kubernetes has become the de facto orchestration technology for managing software containers and is beginning to extend its reach into other areas, including infrastructure management, which The Next Platform spoke about last year. In a report last year, Red Hat said that 88 percent of survey respondents said they were using Kubernetes for container orchestration and 74 percent of those using Kubernetes said they are running workloads in production.
Cisco Systems has embraced Kubernetes for several years, and last year announced the general availability of IKS – or Intersight Kubernetes Service. As the name says, IKS is part of Cisco’s larger Intersight cloud platform, which came into the market more than three years ago.
Cisco, like many of its competitors, sees Kubernetes as foundational to its efforts around the expanding cloud-native software environment that includes containers and microservices and is a key driver of the rapidly evolving IT environment that includes hybrid clouds and edge and demands more flexible and scalable compute, storage and networking capabilities.
For vendors, Kubernetes comes with its own strengths and challenges. According to DD Dasgupta, vice president of product management of Cisco’s Cloud and Compute Business Unit, one of the benefits is that there isn’t one leader in the growing space and the vendor’s strengths in infrastructure – from compute and storage to networking and security – enables it to rethink how they operate in a world where workloads are shifting from just virtual machines (VMs) to more containers on bare metal.
This week, Cisco is stretching the reach of Intersight via IKS, enabling it now to manage Kubernetes clusters and VMs in AWS, Azure and Google Cloud and to manage Kubernetes operations on premises and in the cloud via Intersight.
“Essentially, from an IT standpoint, if you can actually manage all of the Kubernetes clusters, whether they’re on-prem or in a public cloud from a single pane of glass, it becomes a part of a domain that an admin has visibility into and can manage with it,” Dasgupta tells The Next Platform. “Otherwise some developer in a line of business will just go fire up clusters in a public cloud.”
The expansion of Intersight’s capabilities was part of a larger push in both hardware and software Cisco is making to grow its hybrid cloud portfolio. That push includes new capabilities within its HyperFlex hyperconverged infrastructure (HCI) and X-Series modular systems to offer consistent operations and infrastructure and an easier entry point to HyperFlex.
Cisco has more than 30,000 customers using Intersight, which is being used to manage about 415,000 devices worldwide, Dasgupta says. A key is ensuring that enterprises don’t build silos but instead use Intersight to manage its entire IT environment. While it is widely used on premises, it’s also a software-as-a-service (SaaS), so it runs in the cloud. Extending its capabilities to manage Kubernetes clusters in public clouds – and to integrate with AWS’s EC2, where it includes automating VMs in AWS and on premises – is part of that effort.
Basic Kubernetes management is being commoditized, so the value to enterprises is shifting more to having a single tool that can manage all of their Kubernetes deployments, both in the datacenter and the cloud.
“From IT standpoint, you get a common dashboard,” he says. “Even the look and feel, just democratizing and standardizing across that, customers we’ve shown this to like this. You’ve normalized across five different public clouds and you’ll see on-prem versions and you’re providing me this common dashboard. Now I can do automation, governance, inject policy and inspect and because it’s part of Intersight, it comes with all of the classic enterprise tools that IT operators need, like audit and rollbacks, a complete trail of who did what at any given point of time, anything that was executed, that failed or passed. The code on it. You just don’t have tools like this in the industry today that democratizes governance. It is management across all public cloud and on-prem, coupled with all of the enterprise-grade tools.”
That consistency across on-premises and the cloud includes hardware. With HyperFlex, Cisco is rolling out the HyperFlex Express, which includes simplified hardware and software at a price half of other appliances in the family. The vendor also is bringing AMD’s “Milan” EPYC 7003 server chips introduced last year to the HyperFlex systems, a first for the company’s HCI appliances that until now have only offered Intel processors. That said, Cisco does offer other servers powered by AMD chips, including its C-Series systems.
For the X-Series systems, Cisco is now supporting Nvidia GPU nodes via is X-Fabric technology, enabling enterprises to connect compute nodes to the GPUs with the aim of making it easier for organizations to run such modern workloads as AI, machine learning and virtual desktop infrastructure (VDI) and increasing the efficiency in the blade and rack environments.
“You can basically mix and match GPUs with the existing compute nodes,” Dasgupta says. “People have done it in the past with ugly cables hanging out, connecting to a repository of GPUs. But our design … is actually a backplane-less design, so essentially you could have a compute node and a GPU node right next to each other and the whole system shares the available memory drive and GPU capability.”
The vendor also is ramping up the bandwidth in its fifth generation Unified Fabric, an integrated network fabric connecting each X-Series blade or rack system, to 14.84 Tb/sec of aggregate bandwidth, with 100 Gb/sec port support and 200 Gb/sec of bandwidth to a single server. The nature of modern applications, such as AI, demand higher bandwidth and I/O rates, he says.
Cisco announcements have the vibe of the composability message that other vendors like Hewlett Packard Enterprise talk about, with infrastructure and software being easily pulled together based on the demands of an application. Dasgupta says Intersight allows enterprises to have such capabilities, but that is comes from more of a software focus than competitors.
Cisco “comes at it from a software-based set of policies to carve out your existing hardware for the needs of particular applications,” he says. “It provides that granularity that you want. Do you want to run an AI or ML workload or do you want to run it for SAP or do you want to do VDI? You could keep modifying the box based on the profiles that are implemented in Intersight and then it goes and programs the underlying hardware. The GPU nodes in this whole X-Fabric technology are essentially the capabilities of the hardware, but then it’s all going to be micro-segmented and managed by Intersight.”
The vendor’s approach is “not really chasing the whole holy grail of composability, getting into the pedantic definitions of what it is or what it is,” Dasgupta says. “But the admins want to keep leveraging the latest technology from Intel, from AMD, from Nvidia. ‘I want to leverage from memory from this, I want to leverage all of that, but I want to be able to change it at the speed of software.’ Intersight is the overall strategy that we’re executing on.”