Our client is an international automotive supplier with 1750 employees. They currently operate their own central data center and, to provide higher availability, a rented co-location for critical infrastructure workloads. The goal was to find a solution that aligns with the company's new Cloud First strategy while maintaining and increasing availability. The main challenge was ensuring the availability of critical manufacturing systems without the need for application refactoring.
"A stretched cluster configuration is typically used for such high availability solutions," explains Jens Hennig, comdivision's lead architect for this client. He describes stretched clusters as follows: "Stretched clusters expand the vSAN cluster from a single location to two locations, ensuring higher availability and cross-location load balancing. In a stretched cluster configuration, both locations are active. If one site fails, vSAN utilizes the storage in the other site, and vSphere HA restarts any VMs that need to be restarted on the remaining active site."
Hennig explains that the customer wants to increase their already high requirements for the availability of critical systems in the future. "The use of a stretched vSAN not only increases availability but also offers high performance and is future-proof, regardless of whether we want to host virtual machines or containers."A corresponding solution needed to be developed as running a stretched cluster configuration across two locations requires significant design effort if you plan to set up everything yourself.
"...or you can opt for VMware Cloud on AWS," suggests Hennig with a wink. "VMware Cloud on AWS comes with preconfigured stretched cluster functionality, allowing us to distribute stretched clusters across two availability zones and achieve higher availability," Hennig explains. The existing workloads were easily migrated to the cloud.
Additionally, VMware Cloud on AWS reduces costs for the witness host. In addition to the ordered hosts, a witness host with reduced functionality is added to prevent issues such as split brain in the event of a network partition.
"It's important to understand that this is a highly available solution, but not a disaster recovery one," emphasizes Hennig. "The two availability zones remain within a single AWS region, so it does not secure the infrastructure in the event of a regional disaster. Therefore, we advised the customer to use VMware Site Recovery Manager (SRM) to assign critical machines to a different DR target. Initially, this would involve on-premises recovery, but as the on-prem environment gradually decreases worldwide, it can also extend to other VMware Cloud in AWS locations."
Thanks to a quicker implementation without application refactoring or other on-premises infrastructure changes, our customer successfully transitioned to the cloud ahead of schedule. They are currently evaluating the further elimination of physical data centers in two Asian locations.