Edge Recording: Managing Memory

When considering video archiving, the most common scenario is to opt for centralised solution. All video streams from across a site are brought back to a central point, usually a control room, where a recording system stores the relevant data. With recent advances in technology, the concept of distributed archiving has become a reality, with storage devices located at various points of the system. While centralised recording remains the most popular choice, is this for the right reasons? Benchmark considers the options.

[dropcap]C[/dropcap]onsiderations for video archiving used to be simple, predominantly because the archiving technology’s options were limited. There weren’t many choices for recording in video surveillance systems. The decision many made was decided by how much video they were happy to lose, or to put it another way, how much video they could afford to keep.

For many years, centralised recording was – and to a certain degree still is – the predominant model when designing video surveillance solutions. The reason for its popularity is historical. In the past, technology simply didn’t allow any other approach to be achieved for a realistic cost. Every camera needed a dedicated cable linking it to a central point, but recording devices only handled a single video input. The way to record multiple cameras required all inputs being connected to one management unit – typically a switcher or a multiplexer – which then delivered the images from the various devices as one single stream.

It was this limited topology that forced the design of surveillance systems to feature a centralised recording system. This could be a single recorder, or several cascaded units. Even when VCRs were replaced with DVRs, the centralised approach was retained because of the same limitations. DVRs typically included a multiplexer, and so the numerous video feeds still needed to come back to the device to be managed before recording.

As DVRs became the first choice of those recording video data, the idea of centralised recording was challenged for the first time. DVRs could be networked, and this therefore allowed distributed machines to be linked. However, whilst the theory existed, the reality was different.

At the time, most DVRs offered very limited control and management when networked. Also, they were typically used as VCR replacements. In effect, the cabling was already in place for a centralised archiving model, and so the digital units were simply slotted in where analogue machines were being replaced.

Even when digital recording became mainstream, the centralised model remained well established, and for good reason. Legacy cabling meant that the links were already in place to bring back video to a central location for management and archiving. The idea of putting recording devices out in the field was hampered for three main reasons: connectivity, management and cost.

Because VCRs and early DVRs only featured basic remote connectivity, this was a negative for a distributed model. For example, if someone wanted to review high quality footage from a distributed device, an operator would have to visit the recorder, replace the tape or burn a disk, and bring the footage back. It wasn’t an ideal solution. Additionally, this was in a time when surveillance equipment was expensive, and for many the use of multiple recorders was not an option.

There were other factors that made centralised archiving the method of choice. Memory was expensive, and the choices were limited to integral hard drives in video recorders, WORM (write once, read many) media or digital tape from the IT sector. If people thought hard drives were expensive at that time, the cost of digital tape devices and the tapes themselves soon sent them running for cover!

It is true to say that centralised recording model grew out of a need to offset the limitations of certain technologies. This influenced the development of products at that time (and this has carried through to today), and as a result the industry has many systems based around centralised archiving.

No reason?
Putting the cost and connectivity issues to one side, probably the biggest reason that distributed archiving didn’t set the surveillance sector alight in the past was that there was no real need for it.

There weren’t bandwidth issues. Every device used a dedicated coaxial cable, and each link had enough bandwidth to deliver real-time full resolution video. As a result, there weren’t any side-effects to pulling the video back to a central point. It was the accepted (and only) way or working.

Additionally, the topology of systems almost lent itself to a centralised model. The technology simply wasn’t flexible enough to allow systems to be designed in different ways to suit a site’s needs.

Probably the only real negative of a centralised approach is the fact that any system using it has a single point of failure – the recorder itself. To eradicate this, a redundant secondary archiving system needs to be in place. This effectively doubles the archiving cost!

All change
In the last few years the situation has changed dramatically. The introduction of networked-based video surveillance has allowed systems to utilise ever more flexible topologies, delivering economies with regard to cabling and overall design.

Image resolutions are increasing all the time, to a degree that coax-based infrastructure cannot support the footage. These higher resolutions can create heavy demands for bandwidth.

The cost of memory has fallen significantly, and the types of available archiving media have never been so varied. Additionally, the advanced functionality of today’s systems gives the user greater control over the capture and storage of video data. These options not only include how video is recorded, but also how and where the footage is archived. Higher resolutions, increased frame rates and the use of metadata are all in demand, as is the ability to remotely back-up recordings. The end result is an ability to better manage footage, but it places new pressures on the archiving model.

With a networked video surveillance solution, the centralised model has two potential weaknesses. Firstly, if the network becomes congested, or if a link falls over, then the video cannot be archived. Not only is the footage lost, but the recorder will not receive footage until a full service is resumed. Also, if a centralised recorder fails, again all video footage is lost, and new footage cannot be captured until the device is repaired or replaced.

Secondly, if you have a high number of video streams arriving at the central device simultaneously, then bandwidth management is necessary. With a centralised model, this means either accepting poor network performance and its impact on the system, or the overall quality and frame rate of video must be reduced to maximise network performance. The downsides are obvious: the trade-off is a congested network versus poor quality footage.

How do you safeguard against such issues? The distributed model does it by eliminating the dependence upon the weak points in the infrastructure. Dual-streaming cameras and codecs can deliver high resolution real-time video directly to local archiving devices. A second lower resolution bandwidth-friendly stream is then transmitted across the network for viewing.

How you define ‘distributed’ can be decided on a case-by-case basis. In some applications, a group of cameras can be classed as one archiving point. In others, it might be better to archive on a camera-by-camera basis. Thankfully, the wide availability of cost-effective archiving media makes this a reality.

Such distributed archiving points could be analogue, digital or hybrid. This allows selection of the best tools – in terms of cost-efficiency and performance – for every element of the system. It also allows existing infrastructure to be utilised.

The various archiving points could then be linked together over the network, allowing the user to access and manage video as and when they need. High quality video can be downloaded at times when no other network traffic is expected, and if a network link is lost, the video is still securely archived locally.

Equally, if an archiving point fails, and let’s be honest and accept that digital media does fail from time to time, then the rest of the system will carry on working as normal. Also, during any required maintenance, the end user can still use the vast majority of the system, with an awareness of which points are temporarily disabled.

There is another consideration that must be addressed: cost. The assumption that more archiving devices equals increased costs is based upon the traditional centralised approach to recording. If you step away from that, you soon discover that archiving media comes in many shapes and sizes nowadays, and amongst the options are some that are very well suited to the needs of the surveillance sector.

Diverse options
As digital media increases in capacity and decreases in cost – a situation that is increasingly driven by the consumer market-place – the argument for distributed archiving does start to become more attractive. Additionally, the IT sector and multi-media industries have created a growing market for NAS (network attached storage) devices. These offer anything from simple single disk solutions up to multi-disk RAID enabled solutions, and everything in between.

There are a few provisos with regard to selecting a NAS solution. Firstly, ensure that the hardware will be compatible with your systems. Benchmark has seen some NAS manufacturers trying to dump obsolete products into the surveillance sector. They add a paragraph or two about the suitability of the product for surveillance to spec sheets, and then attempt to sell old stock as a security tool. Needless to say, such units haven’t fared well in our reports!

Beware of any NAS product that doesn’t use the latest technologies. If anything relies on serial connections or uses terminal emulators for set-up, you’d be better off avoiding them. If consumers and the IT industry are giving them a wide berth, that tells you something!

Also, remember that some HDDs aren’t designed to have data written to them continually. Always select drives that are rated for AV use, or that are specifically designated for surveillance needs. Solid state drives are also available, and whilst these carry a cost premium they do remove the weak point of moving parts.

Memory cards are also extremely cost-effective. Whilst SD cards are the most common format in surveillance, other options such as Compact Flash are also used. If a reputable brand is selected, these cards are extremely reliable, and capacities are increasing all of the time.

Reliability of devices such as SD cards is high, and with the move to SD-XC, capacities of 128GB and write speeds of 100Mbps are now common. Some people think cards cannot deliver the right capacities for surveillance recording, but a 128GB card for a single device is approximately equivalent to the allocation for a single channel on a 16 channel recorder with 2TB of integral storage.

When selecting SD cards, ensure that you check which formats are supported by the device in question.

One criticism of edge recording is that if the camera is stolen, you not only lose the device, but you also lose the footage! Because the footage is on-board, this is true. However, if the integrator is aware of this it is not difficult to secure the device. Also, edge recording can be carried out on discrete devices, which can be housed nearby in a secure location. Of course, if an NVR is stolen, you lose the footage from every camera! It’s an issue that can be addressed.

Verdict
Centralised recording was born out of the limitations of now redundant technology. Distributed recording is growing in popularity due to the current limitations of an emerging technology.
Both models have their roots in the weaknesses of the technologies employed. This is worth remembering, because evangelists (from both camps) often use the weakness of the alternative model as a stick to beat it with. They then remain very quiet about the weakness of their preferred model!

When you look at the choices, both have redeeming features, and both have possible negatives. Which best suits a site’s needs depends very much on the infrastructure being used, the system’s profile and the needs of the user.