olly - Fotolia

Tip

Hybrid cloud strategies to bridge on-premises, public cloud gap

Moving workloads to a hybrid cloud environment requires several strategies, and organizations should look toward multi-hypervisors, SDDCs and Kubernetes to do so.

Traditional hypervisor vendors have forged new relationships and platforms to extend existing data center operating models into the public cloud. VMware, Nutanix and Microsoft all have unique visions for that integration.

Organizations looking to extend their existing data center workloads to the public cloud or run public cloud workloads on premises should consider hybrid cloud strategies such as implementing multi-hypervisors, software-defined data centers (SDDCs) and/or products such as Azure Stack, Outposts and Kubernetes.

The concept of hypervisor as a commodity

During an interview with Nutanix CEO, Dheeraj Pandey at the Nutanix .NEXT 2018 conference in New Orleans, Pandey explained his idea of the hypervisor becoming a commodity. As organizations and their data centers begin to grow in exponential numbers, Pandey foresees the hypervisor's inevitable transition into the consumer realm.

If admins are looking to incorporate multi-hypervisors into their virtual environments, they should also entertain using traditional hypervisor software to extend legacy applications to the cloud.

"Software provides access  and if you don't have access, you'll always stay at the mercy of the appliance gravity. There [are] a lot of things involved that really keep you anchored to the bottom ... and the only way to unleash this is to really bring more digital delivery models, and software is one such thing," Pandey said.

There are three basic hybrid cloud strategies: SDDCs, cloud-native on-premises tools and Kubernetes.

From traditional data centers to the SDDC

The SDDC is one method to ensure that legacy applications and cloud computing services have support. SDDCs enable organizations to control their data centers through fully automated software, as opposed to traditional data centers that rely solely on hardware and devices.

If admins are looking to incorporate multi-hypervisors into their virtual environments, they should also entertain using traditional hypervisor software to extend legacy applications to the cloud.

For example, VMware has taken advantage of the popularity of vSphere by extending the SDDC construct in the form of VMware Cloud. VMware enables organizations to keep a consistent operating model by deploying vSphere in AWS and Azure. This enables admins to take advantage of traditional vCenter configurations and tools to move applications without modification to the cloud.

Similar to VMware, Nutanix announced the tech preview of its integration with AWS bare-metal Elastic Compute Cloud (EC2) instances. Nutanix enables its Acropolis Operating System platform to operate on a physical cluster running in AWS' cloud. Unlike the VMware-AWS partnership, Nutanix uses generally available services in the AWS data center. Although deep integrations such as extended hyper-converged infrastructure (HCI) storage to Amazon Simple Storage Service don't exist, Nutanix can deploy the software in any AWS region with bare-metal EC2 instances. Organizations can use the Prism management control panel for this software.

Bringing the public cloud workloads on premises


Dheeraj Pandey of Nutanix, sits with Stu Miniman and
Keith Townsend at the Nutanix .NEXT conference 2018
in New Orleans.

Microsoft offers Azure Stack and Azure Stack HCI, which uses hardware to efficiently run cloud workloads on premises. Former Azure Stack chief architect Jeffery Snover explained that Azure Stack is more akin to bringing public cloud into the private data center versus extending Hyper-V management to the public cloud.

Conversely, AWS has announced a comparable option called Outposts. Outposts brings AWS services to on-premises infrastructures such as data centers, co-location spaces and on-premises facilities, which enables organizations to retain their APIs, tools, hardware and functionality.

Kubernetes platforms continue to evolve

Kubernetes offers the ability to abstract the underlying infrastructure and genuinely make the hypervisor a commodity. By migrating workloads to Kubernetes, a cloud control plane -- such as Google Anthos -- becomes the central management plane versus a hypervisor management platform such as vSphere.

But Kubernetes' ability to provide similar levels of management to stateful applications lacks maturity. Google cloud developer advocate Kelsey Hightower agreed: Stateful application deployment in Kubernetes looks tempting, but the execution proves difficult today, he said. Kubernetes itself is only one cloud-native approach to compute, but early Kubernetes adopters believe the shift in development mindset is worth the growth pains in capabilities.

Function as a service (FaaS) products such as Cloud Foundry and Lambda are growing in popularity. These platforms have varying degrees of integration with Kubernetes and VMs, but hypervisor management platforms don't front-end APIs and add little value to these compute models.

VMware has made major investments in the vSphere and Kubernetes integration. In 2018, VMware purchased Kubernetes startup Heptio. With this acquisition, VMware looks to bring VM-based constructs such as NSX-T network overlays, firewalls and NSX Service Mesh to its Kubernetes platform. By doing this, VMware hopes to offer a single approach to managing compute across containers and VMs regardless of the underlying infrastructure.

Nutanix has a competing vision with its Karbon Kubernetes distribution. In the spirit of the company's innovation of HCI, Nutanix looks to bring a simple, one-button management to Kubernetes. However, Nutanix is early in the execution stages of its Kubernetes vision.

Cloud-native platforms such as Kubernetes and FaaS models don't easily translate to hypervisor management platforms, so hybrid cloud strategies should start with cloud-native service management using something other than the hypervisor management platform.

Dig Deeper on Cloud deployment and architecture