April 05, 2016
Agent-Based Or Agentless Backups? Which Is Right For You?
Agent-based or agentless. That is the question. While this isn’t a mission-critical decision, it’s certainly an important one to weigh and determine which is your best fit. There are certain circumstances that call for agent-based backups versus agentless and vice versa. Here’s the rundown.
Agent-based backup products require the user to install a lightweight piece of software on each machine they want to protect. Agent software lives at the kernel level in a protected system so it can easily detect block-level changes on the machine. It doesn't need to scan the entire file system to determine changes for incremental backups. As a result, agent-based can be more efficient than agentless for physical machines.
On the other hand, agent-based backups rely on local compute resources to perform the backup and send it to the backup target. Because of this, the backup process can impact application performance if the protected server does not have the computing power necessary to perform backups alongside production workloads. Generally speaking, this isn’t a big issue as long as the server is right-sized to handle necessary tasks.
Agentless backup, as the name implies, does not require users to install agents on protected servers, therefore making it slightly easier to deploy and monitor. This is particularly important in virtual environments with many production machines in an environment, VMs can be spun up so easily it’s common for them to go unprotected when using agent-based backup because a backup agent is never installed. With agentless, this isn’t an issue.
Also, agentless backups can be less expensive than agent-based backups, because users do not need to pay licensing fees on a per-agent basis. Pricing models, of course, vary from vendor to vendor.
Many agentless backup products, for physical production machines, must inject temporary executable modules into systems before the backup can occur. For the most part, this isn’t an issue, especially if the application is backed up offers an API for the backup software to interface with. However, there is a chance that temporarily injected modules can create data integrity issues on protected machines.
Where agentless backup procedures really shine is VM infrastructure. A good agentless backup software uses the Change Block Tracking (CBT) that exists with the Hypervisor. You get more benefits using agentless backups with a virtual storage location that understands how to transmit data from the hypervisor to the data store without using networking traffic. Therefore, you don't have to use IOPS on the local VM or use network bandwidth that would be better used for actual production processes.