Veeam v13: WAN Accelerator Deployment

WAN acceleration is Veeam's technology for reducing how much data actually travels across the wire during off-site backup copy jobs and replication jobs. It is not a general-purpose network optimizer - it specifically sits in the data path for those two job types and uses global deduplication plus compression to shrink what gets sent over the WAN link.

The mechanism requires a pair of WAN accelerators: one on the source site, one on the target site. The source-side accelerator analyzes outgoing data blocks and compares them against a digest store of previously transferred blocks. Blocks that have already been sent are referenced rather than retransmitted. The target-side accelerator maintains a global cache of received data that all source accelerators feeding into it can draw from. The net effect is that after the first full transfer, subsequent cycles send only genuinely changed data blocks - not just changed blocks relative to the previous job run, but changed blocks relative to everything that accelerator pair has ever transferred for similar workloads.

This is different from incremental backup, which already only captures changed blocks. WAN acceleration adds a second layer: cross-VM, cross-job global deduplication across everything the accelerator pair has processed. In environments where VMs share common OS base images, common application binaries, or common data patterns, the savings can be substantial.

WAN acceleration is worth deploying when your off-site backup copy or replication jobs are constrained by WAN bandwidth, and when your backup data contains a significant amount of similar content across VMs. The more homogeneous your VM population - shared OS images, common application stacks - the better the dedup ratios. In environments with highly diverse or incompressible data, the overhead of running WAN accelerators may not pay off.

The Veeam documentation draws a clear line at 100 Mbps: connections faster than that should use High bandwidth mode, which disables global cache and relies only on cross-restore-point deduplication. For connections below 100 Mbps where WAN capacity is the real constraint, Low bandwidth mode with full global cache is the right choice.

Skip WAN acceleration if: your WAN link is fast enough that backup copy jobs complete within their window without it, your data is highly incompressible (databases with encryption, already-compressed media), or you are using Veeam Cloud Connect where the service provider manages WAN acceleration on their side.

Veeam WAN acceleration operates in two distinct modes, and the mode affects both disk space requirements and what deduplication actually happens.

Low bandwidth mode is the full deduplication path. The source accelerator maintains a digest file for every data block it has ever processed. On each job run, it computes digests for outgoing blocks and compares them against the digest store. Matching blocks are not retransmitted - instead, the target accelerator pulls the block from its global cache. This mode requires the target accelerator to maintain a global cache on disk sized at roughly 10 GB per distinct OS type present across all processed VMs. Source disk requirements scale with the number of VMs and their provisioned size.

High bandwidth mode drops the global cache entirely and performs deduplication only by referencing previously copied restore points that already exist on the target repository. There is no global cache on the target, so no additional disk space is needed beyond the backup data itself. Deduplication ratios are lower than Low bandwidth mode, but so is the processing overhead. Veeam recommends High bandwidth mode for WAN connections faster than 100 Mbps. Both the source and target WAN accelerators must have High bandwidth mode enabled for it to take effect - if either side has it disabled, the pair falls back to Low bandwidth mode.

WAN accelerators can run on Windows or Linux machines. In v13, they can also be deployed on the Veeam Infrastructure Appliance (the same JeOS-based appliance used for backup proxies and repositories), which gives you certificate-based authentication and centrally managed updates. The machine must already be added to the Veeam console as a managed server before you can assign it the WAN accelerator role.

The minimum RAM recommendation is 8 GB. WAN acceleration operations are CPU and memory intensive - the source accelerator in particular requires significant resources during digest computation. Do not put a source WAN accelerator on a machine that is already under heavy load from other roles.

You deploy WAN accelerators through the Backup Infrastructure view in the VBR console. The process is the same for both source and target accelerators - what differs is the machine you point the wizard at and the cache size you configure.

WAN acceleration is enabled per job at the job configuration level, not globally. Both new and existing backup copy jobs can be configured to use accelerators.

The global cache on the target accelerator requires occasional maintenance. Veeam populates and manages the cache automatically during normal operations, but there are two situations where you will need to intervene manually.

The first is cache corruption. If a cache becomes corrupt - which can happen after an unclean shutdown or storage failure - the safest path is to clear it and let Veeam repopulate it from the next job run. Corrupted cache data causes failed or degraded transfers. Clear it rather than troubleshoot it.

The second is when you switch workloads. If you retire a set of VMs and replace them with VMs running entirely different OS types, the existing cache data is no longer useful. Clearing the cache removes data for OS types that no longer exist in your environment, freeing space and allowing the cache to be rebuilt with data relevant to the new workload.

To clear the cache: open Backup Infrastructure, click WAN Accelerators in the inventory pane, right-click the target accelerator, and select Clear cache. The cache is cleared immediately. The next job run repopulates it from scratch, meaning the first run after clearing will have no deduplication benefit on the target side.

A single target WAN accelerator can serve multiple source accelerators from different sites simultaneously. This is the standard architecture for hub-and-spoke deployments where multiple branch offices or remote sites all replicate back to a central data center.

The target accelerator maintains a shared global cache that all source accelerators can benefit from. When source accelerator A sends a block of data, that block goes into the global cache. When source accelerator B later sends the same block - from a different source site but with the same underlying data - it is pulled from cache rather than transmitted again. This cross-site global deduplication is the main advantage of the many-to-one topology.

The practical constraint is that the target accelerator's CPU and RAM must be scaled to handle the concurrent load from all connected source accelerators. For each additional source site, add compute capacity to the target accelerator. Monitor CPU and memory utilization on the target accelerator during peak job windows and scale accordingly.

If you have existing WAN accelerators from a v12 deployment, they carry over to v13 without reconfiguration. The accelerator role remains assigned to the same machines, existing cache data is preserved, and backup copy jobs continue using the same accelerator pair after the VBR upgrade completes.

The new capability in v13 is the Infrastructure Appliance deployment option for WAN accelerators. If you want to migrate an existing WAN accelerator running on a Windows or Linux machine to the JeOS appliance, this requires deploying a new appliance, adding the new WAN accelerator, updating your backup copy jobs to reference the new accelerator, and then removing the old accelerator from the infrastructure. There is no in-place migration from an existing OS install to the appliance - it is a new deployment with job reconfiguration.

For most teams upgrading from v12, leaving the existing WAN accelerators in place on their current machines is the right call. The appliance option is a consideration for new greenfield deployments or when you are already rebuilding accelerator infrastructure for other reasons.

Use this table to match your scenario to the right WAN accelerator configuration.