Bandwidth Fundamentals For Video SurveillanceBy: IPVM Team, Published on Feb 23, 2015
Bandwidth is the most fundamental element of computer networking for video surveillance systems. Because video surveillance can consume an immense amount of bandwidth and because variations in bandwidth load of surveillance cameras can be so significant, understanding bandwidth for video surveillance is critical.
In this in-depth guide, we break down each of the following:
- Measuring Bandwidth
- Bits vs Bytes
- Kilo vs Mega vs Giga
- Bit Rates
- Compression and Bandwidth
- Bandwidth Per Camera
- Constant vs Variable Bit Rates (CBR vs VBR)
- Drivers of Camera Bandwidth Consumption
- Practical Examples of Camera Bandwidth
- Bandwidth Variance Over Time
- Bandwidth and Recorder Placement
- Client Viewing: Multi-Streaming and Transcoding
- Symmetric vs Asymmetric Networks
- Network Bandwidth Capacities
- LAN vs WAN
- Sizing Networks for Video Surveillance
- Quiz Yourself: 10 Question Quiz to measure your knowledge on bandwidth for video networks
Bandwidth is typically measured in bits (e.g., 100Kb/s, 1Mb/s, 1000Mb/s, etc.). A bit is the most fundamental unit of bandwidth and storage.
You should be comfortable measuring the bandwidth, in bits, on your computer. On a PC, this can be done by opening up the task manager as shown below:
On your computer, it typically shows bandwidth being received by and bandwidth being sent out from your computer (i.e., when you watch a YouTube video you are receiving bandwidth, when you send an email you are transmitting bandwidth). These are also known as download and upload speeds respectively.
Bits vs Bytes
In video surveillance, bandwidth is typically measured in bits but sometimes measured in bytes, causing confusion. 8 bits equals 1 byte, so someone saying 40 Megabits per second and another person saying 5 Megabytes per second mean the same thing but is easy to misunderstand or mishear.
Bits and bytes both use the same letter for shorthand reference. The only difference is that bits uses a lower case ‘b’ and bytes uses an upper case ‘B’. You can remember this by recalling that bytes are ‘bigger’ than bits. We see people confuse this often because at a glance they look similar. For example, 100Kb/s and 100KB/s, the latter is 8x greater than the former.
We recommend you use bits when describing video surveillance bandwidth but beware that some people, often from the server / storage side, will use bytes. Because of this, be alert and ask for confirmation if there is any unclarity (i.e., “Sorry did you say X bits or bytes”).
Kilo vs Mega vs Giga
It takes a lot of bits (or bytes) to send video. In practice, you will never have a video stream of 500b/s or even 500B/s. Video generally needs at least thousands or millions of bits. Aggregated video streams often need billions of bits.
The common expression / prefixes for expressing large amount of bandwidth are:
- Kilobits, is thousands, e.g., 500Kb/s is equal to 500,000b/s. An individual video stream in the kilobits tends to be either low resolution or low frame or high compression (or all of the above).
- Megabits is millions, e.g., 5Mb/s is equal to 5,000,000b/s. An individual HD / MP video stream tends to be in the single digit megabits (e.g., 2Mb/s or 4Mb/s or 8Mb/s are fairly common ranges). More than 10Mb/s for an individual video stream is less common (the most typical case is from using the less bandwidth efficient MJPEG codec). However, a 100 cameras being streamed at the same time can routinely require 200Mb/s or 400Mb/s, etc.
- Gigabits is billions, e.g., 5Gb/s is equal to 5,000,000,000b/s. One rarely needs more than a gigabit of bandwidth for video surveillance unless one has a very large-scale surveillance system backhauling all video to a central site.
Bandwidth is like vehicle speed. It is a rate over time. So just like you might say you were driving 60mph (or 96kph), you could say the bandwidth of a camera is 600Kb/s, i.e., that 600 kilobits were transmitted in a second. If you say the bandwidth of your camera is 600Kb or 600KB, not only will you be wrong, you will look incompetent.
Bit rates are always expressed as data (bits or bytes) over a second. Per minute or hour are not applicable, primarily because networking equipment is rated as what the device can handle per second.
Compression and Bandwidth
Essentially all video surveillance that is sent on an IP network is compressed. Surveillance cameras can produce uncompressed video (e.g., analog) but that is almost always compressed before sending over a network. It is theoretically possible to send uncompressed surveillance video over a network but the immense bit rate of even a single stream (1,000Mb/s+) makes it impractical and unjustifiable for almost all.
Bandwidth Per Camera
Bandwidth is typically measured per camera and the amount of bandwidth each camera needs can vary significantly.
One can and should sum / add up the bandwidth needs of each camera on a network to determine total load. For example, if you have 10 cameras on a network and 3 of them use 4Mb/s, 4 of them use 2Mb/s and 3 of them use 1Mb/s, the total load on the network for those 10 cameras would be 23Mb/s.
Constant vs Variable vs Max Bit Rates (CBR vs VBR vs MBR)
The amount of bandwidth a camera needs at any given time to maintain a specific quality level varies over time, sometimes substantially. For example, a camera might need 1Mb/s for an empty school hallway on a Sunday afternoon but might need 4Mb/s for that same spot come Monday morning.
There are three ways to deal with this:
- Variable bit rate (VBR), where the bit rate changes to keep compression at a set level regardless of activity.
- Maximum bit rate (MBR), also called VBR with a cap, where the bit rate changes but no more than a user defined maximum.
- Constant bit rates (CBR), where the bit rate of the camera does not change even if the scene does.
Knowing what type of bit rate control a camera uses is critical, because it impacts bandwidth load significantly. For more, see: CBR vs VBR vs MBR: Surveillance Streaming.
Drivers of Camera Bandwidth Consumption
There is no set standard or even typical camera bandwidth consumption. Using a vehicle example, on a US highway, you can reasonably estimate that almost all cars will drive between 55mph and 85mph.
For video surveillance, some video feeds are as low as 50Kb/s (.05Mb/s) and others are routinely 300 times higher at (15000Kb/s) 15Mb/s.
Here are a few common drivers of camera bandwidth consumption:
- Resolution: everything else equal, the greater the resolution, the greater the bandwidth
- Frame rate: everything else equal, the greater the frame rate, the greater the bandwidth
- Scene complexity: The more activity in the scene (lots of cars and people moving vs no on in the scene), the greater the bandwidth needed.
- Low light: Night time often, but not always, requires more bandwidth due to noise from cameras. See: Testing Bandwidth vs Low Light.
- Model variations: Some models depending on imager or processing can consume far more or less bandwidth.
- Smart Codecs: This is relatively new (developed over the past couple of years), but some cameras even using the same H.264 codec, can intelligently adapt compression for great bandwidth reduction. See: Smart CODEC Guide
Practical Examples of Camera Bandwidth
The following list is an excerpt from IPVM tests of actual bandwidth recording for a variety of cameras:
- CIF 5FPS Office: 50 KB/s
- 720P 10FPS Conference Room: 0.5 Mb/s
- 720P 30FPS Intersection: 4 Mb/s
- 1080P 10FPS Conference Room: 2 Mb/s
- 1080P 30FPS IR On Intersection: 8 Mb/s
- 5MP 15FPs Panoramic Office: 4.5 Mb/s
- 4K 30FPS Intersection: 7 Mb/s
- 4K 10 FPS Night Outdoors: 32 Mb/s
Bandwidth and Recorder Placement
Video surveillance consumes network bandwidth in one of the following 2 typical scenarios:
- Camera / encoder to recorder: Video is generally generated in different devices than they are recorded (e.g., a camera generates the video, a DVR / NVR / VMS server records it). In between, the video needs to be transmitted. If it goes over an IP network (e.g., IP cameras to NVR / VMS), bandwidth is required.
- Recorder to client: Statistically, a very low percentage of video is watched by humans. Often, where the person is watching is on a different device on an IP network than the recorder. For example, the recorder might be in a rack in an IT closet but the viewer (i.e., client) is on a laptop, mobile phone or a monitoring station.
Because of this design, the overwhelming majority of bandwidth needed in surveillance systems is dictated by (1) camera type and (2) the relative placement of cameras and recorders.
In terms of camera type, non IP cameras (NTSC / PAL analog, Analog HD, HD SDI) typically do not consume network bandwidth unless video is being sent to clients as each camera has a cable directly connected to a recorder.
For all camera types, the relative physical placement of the recorder near the camera significantly impacts bandwidth needs. For example, imagine 1000 cameras, with 100 cameras each at 10 buildings on a campus. If each building has a recorder, the peak bandwidth requirements will be ~90% lower than if there is only a single site for recording (i.e., each building recording its own might only need ~200Mb/s network connection compared to ~2Gb/s if they are all being sent back to one building). There are pros and cons to each approach but knowing where you will place recorders has a major impact.
LAN vs WAN
The local area network (LAN) and the wide area network (WAN) are two common acronyms in networking. LAN, as the name implies, refers to networks that are local to a building or campus. By contrast, the WAN, are networks that connect 'widely' across cities, states, countries, etc.
Relatively speaking, bandwidth is cheaper and easier on LANs than WANs.
Network Bandwidth Capacities
In LANs, the three most common network bandwidth capacities are:
- 1,000Mb/s (1 Gig)
- 10,000Mb/s (10 Gig)
In particular, 100Mb/s and 1,000Mb/s connections are quite ordinary for modern networks. For more, see the IP Network Hardware for Surveillance Guide.
Lower than 100Mb/s networks in LANs are relics of the past. They may exist from networks installed many years ago but no one installs LAN networks under 100Mb/s today.
WANs can deliver the same or more bandwidth as the LAN but the costs tend to be significantly higher (in the order of 10 or 100x more expensive per bit) because these networks need to run great distances and across many obstacles. While one certainly could secure a 1 Gig WAN connection, the likelihood that one would do this for surveillance is very low, given the cost this would typically incur.
Symmetric vs Asymmetric Bandwidth
Many WAN networks / connections have asymmetric bandwidth, a problem for remote monitoring or recording of video.
Symmetric bandwidth means the bandwidth is the same ‘up’ and ‘down’, i.e., a link can send the same amount of bandwidth as it can receive (100Mb/s up and 100Mb/s down is a classic example).
Asymmetric bandwidth means the bandwidth up and down are not the same. Specifically, the bandwidth ‘up’ is frequently much lower than the bandwidth ‘down’. This is common in homes and remote offices. These asymetric connections provide sufficient downstream speeds while only providing ~10% of those speeds for upload. The downstream bandwidth might be 10Mb/s or 25Mb/s but the upstream might only be 500 Kb/s or 2Mb/s. In this example, if someone at home wanted to stream a movie (send it downstream from the cloud / Internet), it would not be a problem but if they wanted to upload a movie (or HD surveillance feed), it would be a problem.
The most common asymmetric bandwidth WAN networks are:
- Cable Modem
The main exceptions, those that offer symmetrical bandwidth commonplace, are:
- Telecommunication / telephony networks (e.g., T1s, T3s) but these are fairly expensive and relatively low bit rate (e.g., respectively 1.5Mb/s and 45Mb/s)
- Fiber to the Home (FTTH) / Business (FTTB) are much less expensive than telephony networks and routinely offer 100Mb/s connections. The main limitation is access to such networks. While increasing over the past decade, they tend to be limited to densely populated urban areas.
Sizing Networks for Video Surveillance
Putting this information together, to size a network for video surveillance, you will need to know:
- How much bandwidth each camera consumes, recognizing that wide variations can exist
- How close (or far) the recorder is going to be placed to the cameras connected to it, presuming they need an IP network
- What the bandwidth of those network connections are and what pre-existing load those networks must also support.
For more, see: How to Calculate Surveillance Storage / Bandwidth
See how much you know: Take the 10 Question Bandwidth for Video Networks Quiz