IP Camera Bandwidth and Storage Calculator

$R_v$ is the video bitrate of one stream in Mb/s, and $R_{\max}$ the VBR limit from the camera settings. $G$ is the number of frames between key frames, and $m$ the motion in the scene from 0 to 1: at $m = 0$ only the key frames carry data, at $m = 1$ every frame does. In CBR (constant bit rate) mode, $R_v$ is constant and equal to the set value.

• At the most motion, every frame takes as much as a key frame, that is, the whole bitrate limit; with no motion, only the key frames carry data, and the frames between them are taken as empty.
• Between these extremes, the bitrate rises linearly with the share of the picture that changes; this is an assumption of the model, not a measurement.
• At the same subjective quality, an HEVC (High Efficiency Video Coding, H.265) encoder needs, according to Ohm et al. (2012), “approximately 50% less bit rate on average” than H.264/AVC (Advanced Video Coding); the result concerns reference encoders and test sequences, not cameras.
• The starting values are Hikvision’s approximate recommendations for its cameras; according to the manufacturer, they can be raised by 20 to 30% in more complex scenes. Other cameras may be set lower or higher, so the bitrate from your camera’s settings is the most accurate.
• The codec treats anything that changes the picture between frames as motion in the scene, including night-time noise under infrared, rain, snow and trees moving in the wind.

$B_h$ is the header bytes in an IP packet, $B_v$ the video bytes in a full packet, and $B_t$ the header bytes per packet including the Ethernet frame on the wire (38 B: the preamble and SFD, start frame delimiter; the header; the FCS checksum, frame check sequence; and the interframe gap). $R_h$ is the header bitrate of one stream in Mb/s: the headers of the full packets and, on average, half a packet’s headers for every frame, because the last packet of a frame is not full; for ONVIF metadata, messages per second take the place of $f$.

$R_s$ is the traffic of the $N$ streams of one type, and $R$ the traffic of all streams in Mb/s.

• Packets are up to 1500 B, the Ethernet MTU (maximum transmission unit), and each carries the RTP, transport and IP headers.
• For TCP, we assume the timestamps option (10 B and 2 B of padding); without it, the overhead of a full packet is 0.89 percentage points lower with IPv4 and 0.92 with IPv6.
• We assume full-size packets and one partial packet per frame; cameras that send smaller packets or split a frame into several slices add more headers.
• We leave out the RTP fragmentation header (2 or 3 B per packet), audio and RTCP (RTP Control Protocol) packets; ONVIF metadata counts only when its stream is switched on.
• We count one direction of a full-duplex link, from the cameras to the server; live view pulled directly from the cameras is a separate stream.
• We do not include TCP acknowledgements, the VLAN (virtual local area network) tag, multicast (sending one stream to several receivers) or jumbo frames (larger than 1500 B).

$C$ is the link capacity, $L$ the planned limit, $U$ the link utilisation, and $S$ the spare capacity to the limit. $n_{\mathrm{add}}$ is the number of streams of a type that still fit, and $n_{\mathrm{max}}$ the number that fit on the link with no other streams; $\lfloor x \rfloor$ means rounding down.

• The 70% limit is a design margin, not a standard: VBR (variable bit rate) bitrates have peaks, and key frames cause short bursts of traffic that rarely coincide across all cameras. Set your own value if your network design rules call for a different one.
• $n_{\mathrm{add}}$ and $n_{\mathrm{max}}$ are calculated for one stream type at a time, with the other settings unchanged.

$R_{\mathrm{rec}}$ is the data bitrate of the recorded streams in Mb/s, and $s_r$ the share of the day the cameras record (1 for continuous recording). $M_d$ is the recording storage per day in GB, $D$ the retention in days, $k_r$ the redundancy multiplier, and $M$ the total storage.

• We count only the data of the recorded streams (video, and ONVIF metadata if it is recorded), without network headers.
• The recording bitrate follows from the motion in the scene set in Video Bitrate; a stream in CBR (constant bit rate) mode takes storage in line with its set value.
• When recording on motion, we take the storage as proportional to the share of the day the cameras record.
• Capacity is given in decimal TB (10¹² B); operating systems often show TiB (tebibytes, 2⁴⁰ B), 9.05% less.