Agentless backup enables server backups to be performed without needing to deploy agent software on the host. Instead, backups are taken directly from storage, or in the case of virtual machines, from the hypervisor.
Historically, individual servers were backed up using a local agent or client software. An agent was needed in this instance because, in many cases, the physical storage used by the server was local or taken from a shared storage system not accessible by the data backup software. The agent was responsible for reading the file systems on the host and acting as a data mover, transferring changed -- or all -- data to the backup software.
The agent also tracked changes and was involved in the restore process. Agent-based backups had an overhead; they consumed CPU on the host and transferred data across the network, so many implementations deployed a separate backup network to avoid affecting production data access.
Implementing agentless backup, on the other hand, has a number of benefits.
- There's no software to maintain on each host. This reduces the administrative and compatibility burden, especially in environments with high virtual machine (VM) turnover. There's a large overhead in administration to ensure that backup agents are always kept up to date, as in many cases, the agent software versions are dependent on both the backup software and the host OS release.
- Licensing is simplified. Licenses for agentless backup can be based on backup capacity rather than having to track individual client installations. In environments with high VM turnover, tracking license usage can be a significant burden, and it can result in licenses being underutilized when assigned to zombie or inactive VMs.
- Minimal host and application impact. Backup can be offloaded to the hypervisor or storage array, reducing the impact on the host and application. Modern hypervisors provide interfaces that offer easy access to changed data on VMs. Storage vendors are also opening up their platforms for access to snapshots. One great advantage here is that changed data streams are, in most cases, block-based rather than file-based, providing a high degree of granularity for tracking changes.
- Removal of the backup network. Data doesn't traverse the production or backup network, so networking is simplified. This can result in significant cost savings.
Agentless backup is great for application deployments as VMs on hypervisors and for physical applications -- like databases -- capable of exposing a backup interface. Where possible, agentless backup is a much more operationally preferable platform, and it should be the deployment method of choice for data backups.
Using cloud-to-cloud backup with agentless backup software
Agentless cloud backup can provide Docker protection
Best practices for virtual machine backup and recovery