There are a number of theories about what is acceptable and what isn’t when it comes to bandwidth management. One thing is obvious, however; as the demand for higher resolution video increases, so the pressure to implement ever more effective bandwidth control also rises. Benchmark considers a number of options.
[dropcap]W[/dropcap]hen it comes to IP-enabled video, there are two major benefits on offer. The first is that the technology allows us to break free of the limitations associated with composite video, where resolution was capped. The second is the ability to exploit the flexibility and interoperability of the infrastructure. Both of these are positive benefits, and when considered in isolation they do seem to promise a brave new world where video surveillance can only get better and better and better…
Of course, a healthy dose of reality will soon put things into a less optimistic frame of reference. Every system has finite resources. Even if you throw large amounts of money at the issues, you will still find that there are ceilings! They might take longer to reach, but one truth is that the costs increase significantly for ever smaller hikes in capacity. There will come a point when even the deepest pockets simply won’t be deep enough.
Bandwidth isn’t a negative issue with regard to network-connected video surveillance. Some do try to paint it that way, but they have their own agendas to pursue! However, bandwidth is one aspect of system design that does require careful management.
The first step in gaining an appreciation of how to control bandwidth is to balance an understanding of surveillance with a good grasp of network capacity.
Down the ‘pipe’
Bandwidth is often referred to in a number of ways: network capacity, throughput, load, data traffic, etc.. What it really boils down to is the amount of data being transferred over a period of time. If you have a large amount of data, and it needs to be transferred at a fast rate, then it requires a larger amount of bandwidth. If you reduce the amount of data, or increase the timescale it will be transferred in (or both), then you reduce the requirement for bandwidth. It comes down to a balance of time and data size.
Now consider surveillance. The first issue is that video streams need to arrive in a timely fashion. There is nothing of benefit if an operator becomes aware of an incident after it has happened. Indeed, such an eventuality might even create more problems due to a false sense of security , along with liability issues if a serious event is missed.
Slowing the flow of data to increase the transmission timescale doesn’t offer a credible method of bandwidth management for surveillance. Many network management techniques have evolved with typical data-handling applications in mind. These work on the basis that when a file is updated and saved, it is transmitted from the relevant node. At other times, no data is being moved across the network from that specific node. Back-ups are scheduled for periods when the network isn’t in general use. However, with video surveillance, the nodes – in this case cameras or encoders – are constantly producing files, sometimes at a rate of 25 files every second! These files are continuously transmitted, from multiple nodes. Even a small transmission delay would very quickly grow to the point that the video would be compromised.
Therefore, when considering bandwidth management for video surveillance, the only credible option is to manage the amount of data being transmitted.
Reducing the load
Reducing the amount of data produced by a video surveillance system doesn’t necessarially mean that the information must be ‘thrown away’. There are two approaches. One is to reduce the total amount of data on the network, and the other is to localise the data handling. With the former option, information will inevitably be lost. The latter option allows all of the data to be stored securely, but only low resolution files are streamed. If an operator needs more detail, then and only then will they bring back the bandwidth-heavy footage.
Modern compression techniques have helped to reduce bandwidth requirements. However, there are a few issues with compression. Some manufacturers make claims such as reductions of video frames to less than 10Kb. Such figures are attainable, but the video quality will be badly affected. Given that one of the reasons for moving to networked video is higher resolutions, such an approach makes no sense! As an aside, it should be remembered that file sizes are measured in bytes, while bandwidth is measured in bits. One byte is made up of eight bits.
The standard methods of reducing data file sizes include reducing the frame rate and lower resolutions. There are issues with both approaches. With frame rate reduction, it is important that the end user is aware of the situation. For example, if they have purchased an HD solution, then any reduction in frame rate will mean that the system is technically no longer delivering HD streams! Equally, where the reason for a system upgrade is to achieve enhanced quality, then a resolution reduction may be counter-productive.
With the latest compression algorithms, there is a possibility to reduce the GOV (group of video) length. Also often referred to as GOP (group of pictures), this adjusts the frequency of the capture of full I-Frames. The I-Frames are used as a reference when creating the predictive P-Frames, and are more bandwidth-heavy. Care needs to be taken when adjusting the GOV length, as long periods without a reference frame can lead to some unexpected artefacts which could call the credibility of evidential video into question.
In many applications, a better understanding of how compression works can facilitate reductions in bandwidth requirements. For example, with predictive compression algorithms, an image with noise caused by low light conditions will require more bandwidth. Even in a constant bit-rate system, the image quality will fall due to larger file sizes. This is because the compression engine ‘sees’ the noise as motion. Therefore it uses less of the detail from the reference I-Frame, and captures (and transmits) more data for the P-Frame.
In general, the cleaner, sharper and brighter the viewed scene and resulting image, the less bandwidth will be required to deliver an image of an acceptable quality. Simply applying best practice can deliver real-world bandwidth savings.