What you will learn in this data backup tip: Virtual machine (VM) backup has always been a thorny issues for VMware users. But there have been many data backup and recovery improvements
VM backup in a VMware virtual infrastructure has never been straightforward. This is because most backup administrators don't see a need to change their backup strategy when they move from backing up physical to virtual servers. They implement agent or client software on each VM just as if it were a physical machine. It worked in the physical world, so why wouldn't it work in the virtual world? Well, it does work, but with some caveats.
Because backup software is optimized to back up as many servers/devices as it can in a short a period of time (which makes sense when attempting to optimize for windows of time), it can overwhelm the I/O of a server with multiple VMs. Imagine 10 VMs attempting to be backed up concurrently from the same physical server. Even the latest x86 multicore processors from Intel and AMD will choke.
Then there's the agent/client software running on each of the VMs. Backup software almost always (with some notable exceptions) requires an agent or client piece of software running on the server being protected. This software scans the server for new data at the block or file level and backs it up at the next scheduled backup timeframe. That piece of software is typically touted as being "lite," meaning low resource utilization. The most common resource utilization number thrown around in the industry is approximately 2%. How that number is achieved varies; however, it does not reflect the resource utilization when the agent/client software is actually performing the backup. Then those resources are much higher. Multiply that by the number of VMs and suddenly you have a serious bottleneck in oversubscribed resources.
VMware recognized these backup problems and implemented VMware snapshots that take a point-in-time snapshot of each VM or virtual machine disk file (VMDK) image. Subsequently, VMware integrated Windows VSS with VMDK snapshots for Windows applications making structured applications (SQL server, Exchange, Oracle, SharePoint, etc.), crash-consistent. Next, VMware implemented VMware Consolidated Backup (VCB) that allowed each VMDK snapshot to be mounted on a proxy Windows server that's backed up separately from the VMs themselves (e.g., no agents on the VMs). Unfortunately, it required additional external physical Windows servers and its performance was slow. With the release of vSphere 4.1, VMware has taken a giant step forward in making VM backups easier and more effective than ever before.
VMware vSphere vStorage APIs for Data Protection and Changed Block Tracking
In vSphere, VMware introduced its vStorage APIs for Data Protection (VADP). VADP allows a physical or virtual backup server to tell vSphere to take a VMDK snapshot of a specific VM and back it up directly to the backup server. The backup software may require an agent or client software to run on the vSphere hypervisor, although it doesn't have to do so. No agents or client software is required on the individual VMs.
The vStorage APIs for Data Protection and Changed Block Tracking allow VMs to be backed up simply and without disrupting applications; however, they are only one piece of the backup puzzle.
VADP also goes one step further. In the past, every VMDK snapshot was a full snapshot of the entire VMDK. This made backing up each VMDK snapshot a lengthy process. It also threatened backup windows as VMDKs continuously grow. The VADP in vSphere 4.1 added Changed Block Tracking (CBT). CBT means that each new backed up VMDK snapshot contains only the changed blocks and not the entire VMDK image.
The vStorage APIs for Data Protection and Changed Block Tracking allow VMs to be backed up simply and without disrupting applications; however, they are only one piece of the backup puzzle. They require backup software that utilizes these pieces. VMware itself offers a low-end package called VMware Data Recovery (VDR). VDR is limited to a maximum of 100 VMs and 1 TB datastores. There's no global capability and it doesn't replicate.
The good news is there are many backup vendor products that are considerably more scalable, feature-rich and take full advantage of VADP and CBT. Vendors such as Acronis Inc., Asigra Inc., CommVault Inc., EMC Corp., Hewlett-Packard (HP) Co., IBM Corp., PhD Technologies, Symantec Corp., Veeam Software, Vizioncore (now Quest Software) and a host of others.
Backing up VMs doesn't have to be the massive headache that it has been. VMware is providing new tools and backup vendors are leveraging them. Take a look at your VM backup strategy today and talk with your backup vendor about VADP and CBT if you are not already taking advantage of this easier, faster paradigm.
About this author: Marc Staimer is the founder, senior analyst, and CDS of Dragon Slayer Consulting in Beaverton, OR. The consulting practice of 11 years has focused in the areas of strategic planning, product development, and market development. With more than 28 years of marketing, sales and business experience in infrastructure, storage, server, software, and virtualization, he's considered one of the industry's leading experts. Marc can be reached at firstname.lastname@example.org
This was first published in September 2010