VMware’s Amazonian incarnation drops requirement to run on dedicated hosts

vSphere users can start smaller without paying cluster-bucks

VMware Explore VMware has made a profound change to its partnership with AWS by removing the need to run its wares on dedicated clusters.

VMware Cloud on AWS was the virtualization giant’s first partnership with a hyperscale public cloud and launched in 2017 with a requirement to rent four hosts and sign up for one or three years and spend at least $16,000 a month. The service was also not very elastic: if you wanted more of VMware Cloud on AWS you needed a new cluster. Downscaling was not an option.

Some tweaks to the service have emerged, such as the 2020 offer to reduce the size of the required cluster from four to two hosts. A couple of extra host options also emerged. But the service retained an un-cloud-like inflexibility as the required cluster configs did not offer an enormous menu of options.

The reason for that not-very-cloudy arrangement was that VMware’s Amazonian incarnation needed to be co-engineered with AWS, not just offered as a cloudy version of existing offerings. VMware was given privileged access to the innards of AWS so that its own abstractions and Amazon’s didn’t clash.

The virtualization giant has admitted that its AWS offering was therefore hard to love – and afford – for plenty of its users, especially its smaller customers.

At the VMware Explore conference this week, the company announced it has freed the service from the requirement to run in a cluster with an offering called “Cloud Flex Compute.”

The offering means users can acquire what was described to The Register as “a slice” of AWS comprising user-defined levels of compute, memory, and storage resources. Each slice will be a VMware environment in which users can create and run the VMs they want, on AWS. But the slice won’t be defined as a host, instead appearing as a pool of resources.

Slices will be quite flexible: VMware told The Register configurations will allow general purpose compute, memory optimized VMs, or storage-centric machines.

Yet vAdmins will manage the slice as if it were any other pool of resources in a private cloud and a vanilla vSphere implementation.

Krish Prasad, VMware’s senior vice president and general manager of cloud platforms, told The Register that over time slices will match different EC2 instance types.

“Over time, you will see a big array of new instance types. There will be small ones and large ones – the entire spectrum.”

That’s significant because slices are elastic: users can increase the resources available

Prasad said customers have asked for smaller, more flexible, VMware on AWS for some time, but the engineering effort required to deliver it was not straightforward.

“It’s the next level of architecture work that we had to do to because if you are providing a single host and the customer is staying in the host, you know, you're secure by default” – or as secure as the isolation provided by a hypervisor allows.

“We had to do additional work, when you start slicing to make sure the isolation is there,” Prasad added.

But slices and VMware Cloud Flex Compute are significant for more than offering smaller, cheaper, access to AWS. They also matter to both companies because the cluster-based service produced rigs that were not right-sized to many applications, meaning VMware on AWS was a poor application migration or disaster recovery option.

Cloud Flex Compute is currently in preview. Once it becomes formally available, VMware on AWS will be far more suitable, and affordable, for many more vSphere users. Maybe even for you. ®