Cohesity running alongside Cisco HyperFlex within a Cisco UCS domain offers a consolidated system that provides the primary storage, workload hosting, data protection, and file services required for most virtualized data centers, all within a single unified architecture. Cohesity and Cisco HyperFlex share complementary data center technologies, both utilizing a distributed file system architecture that is designed for high availability. Through a shared-nothing topology, there is no single point of failure, and there are no inherent bottlenecks, and both performance and capacity can scale linearly as more physical nodes are added to the clusters. The distributed file system spans all nodes in the cluster and natively provides global deduplication, compression, and encryption.
With Cisco HyperFlex integration, Cohesity DataProtect software takes virtual machine snapshots directly on HyperFlex, which creates a storage-native snapshot for the virtual machine. Because this snapshot is native to HyperFlex, it has very similar performance characteristics as that of the original base disk when compared to the performance when using standard VMware redo-log-based snapshots. After a snapshot is taken, Cohesity DataProtect proceeds to back up the virtual machine data, and then the snapshot is deleted through the HyperFlex API. Using native snapshots eliminates common delays and I/O penalties and improves application performance by using the underlying HyperFlex distributed storage technology to create and consolidate the snapshots.
During a Cohesity Protection job, a new snapshot of the virtual machine is taken, and that snapshot is transferred via the network to the storage domain configured in the job. This constitutes a new incremental backup of that virtual machine. Once the snapshot is transferred, the snapshot of the virtual machine is deleted in the source hypervisor node. If the virtual machine being backed up was already running with an active snapshot, the new snapshot taken by Cohesity is a child of the existing snap, and then it is deleted, coalescing the changes back into the existing snapshot level where the virtual machine was already running. If storage snapshot provider integration with Cisco HyperFlex is enabled, then all of these snapshots are taken as HX native snapshots. If the HX native snapshot attempt fails, such as when an existing VMware standard redo-log snapshot exists, the protection job falls back to taking a standard VMware snapshot.
A recovery job can be initiated to restore a virtual machine from the backed-up snapshots and return the virtual machine to service. A unique aspect of the Cohesity software is the sequence of the recovery process. When a recovery job is started, the Cohesity system presents an NFS-based datastore from itself, which is mounted to the ESXi host, inside of which are the virtual machine files that have been bloomed from the snapshots. The virtual machine is then registered in vCenter from this location, and the virtual machine is powered on. This process returns the recovered virtual machine to service much faster than would a typical recovery process because the virtual machine immediately runs with its virtual files sourced from the Cohesity NFS datastore. After the virtual machine is powered on, a storage vMotion relocates the virtual machine files to their original location. The benefit of this recovery workflow is amplified when multiple simultaneous virtual machine recoveries are needed because the time to return the virtual machines to service is very low, and the remaining process of relocating the virtual machines via storage vMotion happens in the background while the virtual machines are already online. A recovered virtual machine has no snapshots, even if the virtual machine originally had snapshots at the time of the backup that is being restored.