VMworld 2010 Reporting: SRM Futures: Host Based Replication

Note: In the next major release fail back will be supported. During fail back only the vms that were failed over initially will be failed back.

Site Recovery Manager of Today

- Simplified and automated testing and failover.

Site Recovery Manager of the future

Host Based Replication (HBR)
HBR provides replication between different storage vendors and local storage. Replication is managed as a property of the virtual machine or a large group. Replication is being managed at the ESX layer. You can opt to replicate all disks or a subset of disks for a vm. HBR Allows you to sneaker net in order to jumpstart replication. You set the RPO (Recovery Point Objective) on a per vm basis down to 15 minutes. HBR is based on replication of delta disks between sites.

ESX watches the lower level SCSI traffic and sends the changes. A FT enabled vm will not be supported when the product ships. The product is expected to ship next year.

There is no log management of applications inside the VMs.

The framework for HBR involves an HBR agent installed on the ESX host on the primary site. A group snapshot of all disks associated with a VM is done to ensure crash consistency across the entire vm. The second component is an HBR server (HBR Management Server) which is a Linux based vm deployed at the replica site. The vm is controlled through the SRM tab, no need to go to the interface of the HMS. Multiple HBR servers can be deployed for scale but one will be typical for most customers. One HMS per vCenter. The HMS has a database (internal or external TBD) to keep track of the linkages. Many to one site replication scenarios are supported. The first release supports powered on vms; replication stops on powered off vms. Physical RDMs are not supported for HBR because it ESX needs to see the SCSI transactions.




- Posted using BlogPress from my iPad