Resiliency VMware Cloud on AWS

4-Host Entry Scale Stretched Clusters

VMware Cloud on AWS has proven its ability to simplify operations and unblock customer success around the world. One of the more critical technologies enabling this shift has been our Multi-AZ resiliency powered by Stretched Clusters. Today we’re excited to share a new entry into this space. Entry scale Stretched Clusters allow a Multi-AZ deployment to be created with as few as 4-Hosts!

Why Stretched Clusters?

When migrating from a traditional data center environment to a AWS public cloud, several critical architectural elements must be understood and respected. The main being Availability Zones. A Region in AWS is constructed of multiple independent Availability Zones (AZ). These islands of infrastructure are entirely self-sufficient and allow an application to maintain availability in the face of catastrophic failure. Per AWS best practices, an application must be distributed across multiple Availability Zones to be truly highly available.

A Stretched Cluster within VMware Cloud on AWS accomplishes this at the infrastructure level by physically stretching the vSphere Clusters across multiple Availability Zones. With vSphere hosts in multiple AZ’s, vSphere HA can recover from an AZ failure without any direct intervention by simply restarting any failed VMs in the surviving AZ. Thus, Stretched Clusters enable customers to move business-critical workloads into VMware Cloud on AWS without the need to refactor, rearchitect, or re-platform their workloads.

What is an Entry Scale Stretched Cluster?

A Stretched Cluster requires six or more hosts. This is to deliver on the stated availability guarantee of 99.99%. To ensure the cluster can survive a combination of failures, we require resiliency within and between each AZ.

Stretched Clusters with 4-hosts do not have enough hosts to maintain vSAN fault tolerance inside the AZ as well as between AZs. They can, however, support synchronous mirroring between AZ’s. This creates a cluster resilient to the loss of an AZ but not as resilient as our existing Stretched Cluster offer. Therefore, we have updated our SLA to introduce a 99.9% availability guarantee for Stretched Clusters with 4-host or less.

When to use a 4-host Stretched Cluster?

4-Host entry scale Stretched Clusters provide additional resiliency above and beyond a traditional Single-AZ SDDC. This is because the SLA carries no dependency on the underlying AZ. A Single-AZ non-stretched cluster offers a 99.9% availability guarantee, but only as long as the underlying AZ is available. If the AZ itself encounters an outage, the service would be unavailable until AWS restores the AZ.

A 4-host Stretched cluster has no such restriction. Even in the event of a complete AZ failure, we guarantee 99.9% availability. Furthermore, an entry scale stretched cluster is still a stretched cluster. Any 4-host stretched cluster can obtain the additional ‘9’ at any point by adding a pair of hosts and updating the storage policy to match the 99.99% SLA requirements.

Considerations

It is important to note that 4-host stretched clusters are currently a starting configuration. The service supports scale-out on 4-host using either the UI/API and or Elastic DRS. However, that scale event is one way. A 6-host stretched cluster cannot be currently scaled back down to a 4-host stretched cluster.

Admission Control has also been modified on any 4-host stretched cluster deployment. In order to guarantee management recovery, Admission Control is configured for 2 host failure toleration. This means that any VM with a reservation will have that reservation doubled. Effectively reserving the resources in both AZs. For this reason, we recommend not configuring any reservations on a 4-host Stretched Cluster.

Summary

4-host Stretched Clusters enable customers to realize Multi-AZ resiliency without significantly increasing the resources requirements. If you’ve been considering VMware Cloud on AWS but found Single-AZ too risky and Multi-AZ too costly, then 4-Host Stretched Clusters maybe your just-right solution.

If you would like to learn more about VMware Cloud on AWS, please check out the resources below: