Nutanix is a hyper-converged infrastructure solution where each node is a self-sufficient set of compute and storage resources to run workloads. Four Nutanix nodes form a Nutanix Block. Multiple Nutanix Blocks can be acquired for pooling compute and storage. Nutanix follows ‘shared nothing’ architecture for storage. While storage from multiple nodes are pooled and presented as a single name space, data locality is ensured as much as possible with the use of a service VM on each node that keeps hot data on server flash and moves cold data to hard drives. Making data copies on other nodes in the cluster ensures availability. For a simple visual on Nutanix architecture, I recommend watching this video
One of the most popular use cases of Nutanix is to run VMware vSphere. How do you backup workloads on VMware vSphere if they are running on Nutanix platform?
Just like any other platform for hosting VMware vSphere, the workloads in Nutanix can be backed up in a number of ways. The old school guest agent based backups as well as the image-based backups using vStorage APIs for Data Protection would work here.
Now let us ask this question. What is the most scalable backup method for vSphere on Nutanix? Does it also deliver performance without competing for compute and I/O from production applications?
The key to performance and scalability in backups is to understand how data from production storage is read during the process. As you guys already know, there are three distinct data transport methods to move data from production data stores to backup storage. SAN Transport, Network Block Device (NBD) transport and hot-add transport.
Nutanix is all about achieving operational simplicity by eliminating expensive storage arrays and complex storage area networks (SAN). Since there is no shared storage or SAN, you cannot use SAN transport method to move data.
Transport using hot-add VMs (sometimes known as ‘Proxy VMs’) is possible on Nutanix platform but not recommended. As you may already know, hot-add VMs leeches resources from production ESXi hosts. They also add complexity and increase operational overhead. Furthermore, on Nutanix, hot-add VM can tax the service VMs on nodes heavily because hot-add VM on one node could be backing up production VMs from other nodes. Nutanix would need to read data from node to node that is sub-optimal. Note that data is not needed anymore after the backup (unlike a vSphere HA case where VM itself is relocated to target Nutanix node) and hence causing unnecessary overhead.You can workaround by keeping a hot-add VM on each node; but that is cumbersome, costly and inefficient use of resources.
By now, you have already figured out the answer. The best way is NBD! You are almost there….
ESXi is VMware’s proprietary platform. In order to present Nutanix’s scale-out storage with all its benefits (server flash and distributed DAS with single name space), our friends in Nutanix cleverly virtualize storage and presents to compute nodes (ESXi hosts) over NFS. For an ESXi host, the traffic for NFS and NBD uses VMkernel ports. Hence this is a perfect marriage to protect VMware vSphere on Nutanix. Furthermore, NBD eliminates the need for complex hot-add VMs and unnecessary internode traffic.
Regular readers of Netting out NetBackup blogs would know that NBD comes with a performance penalty and NetBackup engineering had provided innovations eliminate that problem altogether. All VMware vSphere backup solutions in the market tend to prefer SAN or hot-add over NBD because there is VMkernel enforced bandwidth cap for streaming backups. This is to protect VMkernel from destabilizing if backups use all the available traffic. Unfortunately, this also limits the use of NBD for large enterprise environments. However you have nothing to worry, you got NetBackup!
The NetBackup solution for Nutanix
What you need to do is to make use of a few powerful capabilities in NetBackup.
- NetBackup 7.6 media server or NetBackup Integrated Appliance. NetBackup Integrated Appliance is recommended as you went for Nutanix for its integrated appliance form factor anyway! Choose a multi-10Gb Ethernet model unless you have non-Nutanix environments that can make use of SAN connections.
- Enable VMware Intelligent Policy and set Resource Limits to operate up to 4 streams per ESXi host. We have industry benchmarks to prove that NBD has diminishing returns over 4 streams as it hits the VMkernel enforced bandwidth cap. As VMware Intelligent Policies dynamically throttle streams from ESXi and distributes over the entire set of hosts, you get the best performance from distributed storage architecture in Nutanix. Furthermore, this method guaranties data locality and hence could also retrieve hot data right from server flash in Nutanix node!
- Enable NetBackup Accelerator for VMware: This takes performance and scalability to a whole new level. While NetBackup knows the best way to utilize NBD (hence already the best solution for Nutanix), it goes further with incremental forever strategy with Symantec V-Ray visibility. I have covered the internals of NetBackup Accelerator already in this blog
- Superior performance during initial seeding with the least impact on Nutanix internode network
- Dynamic discovery and throttling across ESXi hosts (Nutanix nodes)
- Data locality guaranteed and hence hot data always retrieved from server flash
- Incremental forever after the initial seed. Near zero impact on Nutanix compute nodes although NetBackup is using NBD for transport
- Unmatched scalability. The more Nutanix nodes you add, the more this backup methodology scales out dynamically across nodes!