Concentrate on HyperFlex: Cloud-driven, many-to-one Catastrophe Restoration


This weblog collection focuses on varied points of Cisco HyperFlex.  In at present’s weblog we’ll take a look at two distinctly totally different catastrophe restoration options and why it’s vital to grasp the use instances underpinning every answer. I’ll speak in regards to the enterprise challenges we try to resolve, and which answer is the very best match.

A quick historical past of HyperFlex Catastrophe Restoration

HyperFlex was initially launched with out native replication — due to this fact native DR — performance. A short while later, in 2017 HyperFlex Knowledge Platform (HXDP) model 2.5 launched help for asynchronous replication of VM snapshots. The native replication topology consists of a 1:1 configuration with two paired HyperFlex clusters. The paired clusters could be geographically separated and make the most of WAN connectivity. Workflow help for VM failover & re-protect, deliberate migration, and test-recovery make it match for customers on the lookout for a single-cluster catastrophe restoration answer, or an active-active answer with every cluster in a position to act as a restoration cluster for the opposite. If you’re considering this sounds a bit like a built-in function with SRM-like capabilities, you’re right!

The most important limitation of that providing is that it isn’t geared in the direction of defending greater than two clusters. Every further protected cluster requires a “companion” cluster to create a cluster pair. Whereas attainable, scaling the 1:1 replication answer comes with some further prices.

Use instances for the 1:1 replication answer embody:

  • Restoration Protected VMs could be recovered on the paired vacation spot cluster within the occasion of a supply cluster outage. The consumer can re-protect the recovered VMs on the level when the unique supply cluster outage has been resolved.
  • Migration – Protected VMs can endure deliberate migration, the place a brand new snapshot is taken and replicated to the paired cluster. The replication route is robotically reversed on the level when migration has accomplished.
  • Testing – Protected VMs could be test-recovered with out impacting manufacturing workloads. This makes the testing of restoration procedures attainable at no matter frequency is required.

The 1:1 restoration answer is managed utilizing the HX Join consumer interface. Entry to HX Join on every of the paired clusters is important when deploying the answer.

Enter: Many-to-One Catastrophe Restoration

Quick ahead to 2021, when new replication expertise was launched with HXDP model 4.5(2a) then referred to as N:1 Replication for Edge. HyperFlex now supported a fan-in replication topology, and that the answer was managed with a cloud-based management airplane named Intersight. The Intersight consumer interface lets customers handle massive numbers of HyperFlex clusters from a single console without having to login to a number of totally different consumer interfaces.

N:1 catastrophe restoration addresses the scalability limitations of the 1:1 answer, nevertheless there may be different much less apparent performance that stands out. Right here’s a fast listing of just a few:

  • Protected Datastore – Any VMs residing within the protected datastore are robotically protected utilizing a standard backup coverage. Customers don’t want to use backup settings to VMs individually, simplifying the administration of the answer.
  • Retaining a number of restoration factors – On each the originating supply cluster in addition to on the vacation spot backup goal cluster customers can choose to retain from 1 to 30 snapshot-based restoration factors. This permits the consumer to get better VMs from a cut-off date earlier than a logical corruption or unintended deletion occasion might have occurred.
  • Restoration of VMs – Both to the unique supply cluster, or to a distinct supply cluster.
    • Recovering a VM on the originating supply cluster leverages a regionally retained snapshot-based restoration level. Restoration operations are very quick as there isn’t a knowledge switch concerned.
    • VMs can optionally be recovered onto a distinct cluster. When performing this restoration operation, the chosen snapshot-based restoration level is replicated in compressed format to the specified cluster. This makes it attainable to get better VMs on a distinct cluster in conditions the place the unique supply cluster could also be unavailable. One other use case is ad-hoc migration of VMs from one supply cluster to a distinct supply cluster.

Not only for the sting anymore

When HXDP model 5.0(1a) grew to become out there in early 2022, N:1 restoration for Edge answer morphed into the Many-to-One (N:1) Catastrophe Restoration for HyperFlex answer. This grew to become attainable when help for FI (Material Interconnect) primarily based supply clusters was added. What was initially a safety answer for distant HyperFlex Edge clusters grew to become an enhanced answer for all HyperFlex clusters. One other important enhancement added at the moment was the power to retain totally different numbers of snapshot-based restoration factors on supply and backup goal clusters. Customers can now decide to retain fewer native restoration factors on a supply cluster and retain a better variety of restoration factors on a backup goal cluster. Fewer restoration factors on a supply cluster supplies a capability to eat much less space for storing when fewer native restoration factors are required.

HXDP model 5.0(1b) added help for software program encryption in order that the protected datastore on a supply cluster can optionally be encrypted. And sure, the reproduction datastore on the backup goal datastore (created robotically) will even be encrypted!

It’s vital to notice that whereas each the 1-to1 and many-to-one replication answer are supported, they can’t be co-mingled on the identical HyperFlex cluster. Value noting is the persevering with effort and onerous work that takes place behind the scenes to enhance and improve HyperFlex replication. Can we count on ongoing enhancements in future releases? That’s a reasonably protected guess!

Hopefully this has supplied a view into the 2 totally different HyperFlex replication options.  For added data, see our assets part beneath.



N:1 Replication set up and troubleshooting information

All issues HyperFlex (touchdown web page)



Please enter your comment!
Please enter your name here