A recording server is the machine in a VMS that pulls camera streams and writes them to storage — the workhorse that does the actual recording. It receives the encoded video from a set of cameras, writes it to disk (often a RAID array or attached storage), and serves it back for live view and playback. A large system runs several recording servers, each responsible for a group of cameras, coordinated by the VMS management layer.
The crucial sizing principle is that a recording server is bounded by sustained write throughput, not by a headline camera count. What it can handle is the total megabits per second of all its cameras' streams, written continuously to storage — so two hundred high-bitrate 4K cameras may be a heavier load than four hundred low-bitrate ones. With enough throughput a single server can record on the order of several hundred cameras (roughly a few gigabits per second of sustained write), but the limit is the data rate, not the device count.
The pitfalls are sizing by camera count and ignoring the write path. A server speced for "300 cameras" can fail if those cameras stream more bitrate than its disks can absorb, or if the RAID write penalty was not counted, or if decode for re-streaming was overlooked. And a recording server is a failure domain: when it goes down, its cameras stop recording unless failover or edge recording covers them. Size by realistic sustained bitrate with headroom, account for the RAID write penalty, and pair recording servers with failover and edge recording for the cameras that must never miss.
