When Veeam sends object storage traffic through a gateway server, every byte for that repository converges on one host. Gateway server placement, interfaces and memory then set the throughput ceiling, often well before the storage does. A repository that reads as slow while the storage behind it looks idle is usually a routing problem, not a storage problem.
The gateway is easy to overlook because it is rarely a dedicated machine. It is a role on a server that already exists, often accepted by default during setup and never revisited.
Veeam can send object traffic either through a designated gateway server or directly from the component doing the work. Which applies depends on the object storage repository configuration and the job type, and the behavior has changed across versions, so the setting in the deployment at hand is worth reading rather than assuming.
Direct traffic scales with the number of proxies and agents. Gateway traffic scales with what one server can carry. A gateway still earns its place when the storage should not be reachable from every proxy, when firewall rules are simpler with one source address, or when certificates and credentials should sit in one place. Direct paths win when throughput comes first, a trade covered in direct to object configurations.
A gateway terminates the object protocol, assembles and sends objects, and handles transport security. The work per byte is light compared with a proxy, but it is continuous and grows with the number of concurrent tasks pointed at the repository. Network interface capacity usually runs out first.
Memory runs out second. Object uploads are buffered, and each concurrent task holds working memory, so a gateway comfortable with a handful of tasks can start failing when asked to carry many. When the role shares a host with the backup console or a proxy, those demands stack on what the server was already doing, and the contention is not reported as a gateway problem. Gateway capacity and task slot counts are therefore one question asked two ways: raising concurrency settings without knowing what the gateway can absorb converts a throughput limit into retries.
Backup object storage is often placed behind a boundary on purpose, in its own segment reachable only through specific rules. The gateway's position relative to that boundary decides whether traffic crosses it once, twice or not at all. A gateway on the production side pulls every byte through the firewall or inspection device on the way to the storage. A gateway on the storage side moves that crossing to one controlled hop and keeps the bulk transfer on the storage network. Inspection devices are sized for user traffic, not a backup window, and are seldom instrumented to show the saturation. Where the segment exists for isolation, placement has to respect what the network segmentation was built to achieve, since moving a gateway across the boundary can undo it.
| Observation | Likely cause | What to check |
|---|---|---|
| Throughput flat well below the link rate | A single gateway interface is saturated | Interface utilization on the gateway during the job, not on the storage |
| Storage nodes show low utilization while jobs run | Traffic is constrained before it reaches the storage | Every device in the path from gateway to storage subnet |
| Throughput falls when a second job starts | CPU or memory contention on the gateway host | Concurrent task count and what else runs on that server |
| Backup acceptable, restore slow, same repository | A different component handled the restore path | Which server performed the restore and which interface it used |
| Traffic crossing a site or segment boundary unexpectedly | Gateway placed on the wrong side of that boundary | Gateway host location relative to the storage subnet and the routing between them |
| Timeouts and retries under load, no errors on the storage | An inspection or proxy device in the path | Devices terminating or inspecting object traffic on that route |
Start at the interface counters on the gateway host during a busy job. If one interface is at line rate, the answer is found and the question becomes how to add paths. If not, check CPU and memory on the same host, then the concurrent task count, then the devices between gateway and storage subnet. Veeam's bottleneck statistics in the job session show which stage the software itself considers the constraint.
Check the storage last on purpose. It is the most common first guess and the least common cause, because object storage spreads writes across many nodes and can usually absorb far more than one gateway sends. Low node utilization while the repository is reported as the bottleneck points to the path. Measure restore separately, since the component that reads is not always the one that wrote, and restore throughput can hit a limit backup testing never touched. Gateway metrics belong beside repository counters in the monitoring the operations team already watches.
Keep one dated page per repository naming the gateway host, its interface and subnet, and what sits between it and the storage, and recheck it after any network or host change. A gateway role can move when a server is rebuilt.