Monday, November 23, 2015

Backing up VMs on VSAN with VDP and Data Domain

Virtual SAN (VSAN) is VMware’s software-defined storage solution for a hyper-converged infrastructure (HCI). Virtual SAN is natively integrated with vSphere and it is configured with just a few mouse clicks. Since disks internal to the vSphere hosts are used to create a Virtual SAN datastore, there is no dependency on external shared storage. Virtual machines can be assigned specific storage policies based on the needs of the applications. However, one aspect that is not (yet) part of a Virtual SAN storage policy is backup and recovery. That is where vSphere Data Protection (VDP) can play an important role. VDP 6.1.1 features interoperability with VSAN 5.5, 6.0, and 6.1 and it is possible to deploy a VDP virtual appliance to a VSAN datastore to back up VMs running on VSAN. However, this breaks the cardinal rule of not having your production data and backup data on the same storage. How can we address that? Keep reading...

First, I will point out that it is not necessarily a bad thing to have production data and backup data living in together on the same storage platform (VSAN, in this case) as long as you have a another copy of this backup data located elsewhere. VSAN is quite resilient to issues such as the failure of a drive or a vSphere host, but that will not help when someone accidentally hits the Delete button. Having your backup data locally available enables rapid restores for situations such as data corruption inside of a VM or that inadvertent deletion of a VM.

VDP can also protect against more significant failures such as the loss of an entire site. The first way this can be addressed is by using the backup data replication feature that is built into VDP. It is possible to back up data locally and then replicate this backup data to another VDP appliance.

It is also possible to replicate backup data from VDP to EMC Avamar. Avamar could be located in the customer’s data center or hosted at a service provider as part of a backup data hosting offering. If multi-tenancy is required, VDP supports it as discussed here.

Another option is to utilize EMC Data Domain as a backup data target for VDP. Data Domain Boost (DDBoost) is built into VDP and, assuming DDBoost is licensed on the Data Domain appliance, this enables a very efficient and reliable means to store backup data separate from production data. Data Domain deduplication and compression is outstanding. Below is a screen shot showing pre- and post-compression numbers for two VDP appliances I have in my environment. Note that over 2TB of raw backup data has been ingested by the two VDP appliances. After compression, only 127GB of capacity on the Data Domain appliances is consumed.

Better yet, what if we employ both approaches? It is certainly supported and looks something like the diagram below. The virtual appliances are VDP appliances and the black boxes are EMC Data Domain (physical) appliances.

You might be asking "What happens if I lose the VDP appliance?" VDP includes the option "Enable Checkpoint Copy", which enables a new VDP appliance to restore existing backup data from a Data Domain system if the original VDP appliance is lost.

As you can see there are a few ways to design a backup solution using VDP and Data Domain for VSAN (and other storage solutions, for that matter). Using Data Domain to store VDP backup data makes sense for a few reasons:
  • Backup data is separate from production data
  • Deduplication and compression reduce the storage capacity consumed by backup data
  • Data Domain provides a very reliable platform for housing backup data
  • VDP works seamlessly with Data Domain using DDBoost
  • Data Domain enables scale beyond the 8TB limit imposed by a VDP appliance
It is possible to use more than one VDP virtual appliance with a single Data Domain appliance. If this is something you are thinking about doing, here are some guidelines to keep in mind:
  • VDP is still limited to a supported maximum of 400 VMs per VDP appliance
  • 1 VDP appliance per DD160, DD620, and DD2200
  • Up to 2 VDP appliances per DD2500 and DD4xxx
  • Up to 3 VDP appliances per DD7200 and DD990
Generally speaking, the number of VMs that can easily by protected by VDP is anywhere from just a few up to about 500 hundred VMs. Scaling to 300 or 400 VMs, for example, with VDP would likely be accomplished by deploying two or three VDP appliances with 8TB of capacity and/or use of Data Domain as a target for backup data. It is important to understand there are several factors that determine actual results such as the amount of data being backed up (VM sizes), data change rates, the types of data backed up, and the retention policy assigned to the backup data. Everyone’s environments are different – results will vary.

Follow me on Twitter: @jhuntervmware

No comments:

Post a Comment