The growing popularity of video surveillance is fuelled by the shift to IP video systems - they are less expensive than traditional analogue systems, and make it easier to add or relocate cameras and multicast video signals to multiple locations.

Yet, every system is limited by two finite resources: network bandwidth and disk space.

A 100Mbit/s connection can support nearly 85 cameras producing 10-15KB images per second (about 1Mbit/s each), but the network will be running at 100 percent utilisation. Whether the IP video system is on the existing IT network or a parallel network, enterprises need to set a realistic bandwidth target and then determine how many cameras can be supported by their infrastructure.

To ensure efficient bandwidth usage, security integrators can help design fluid systems that adjust to accommodate traffic surges. For example, frame rates can be reduced temporarily while the sharpness of each image is maintained, or the same number of frames can be recorded at a lower resolution.

Video content analysis (VCA) embedded in an IP encoder — the device that translates analogue camera signals into digital — can also reduce the amount of traffic sent across the network. With VCA at the edge, it is possible to only transmit video that generates an alarm due to a security concern — such as a person loitering, theft of an object or an object left behind in a scene. By selecting specific portions of video to forward, VCA diminishes the amount of bandwidth required for surveillance but enables all camera channels to be monitored effectively.

Of course, storing all that IP video represents a challenge. A 200GB hard drive can store approximately two weeks of video data, depending on the quality of video. For some industries, such as prisons and gaming, regulations may require that recorded video is stored for a longer period of time — up to 30 days, or even a year in some cases.

Enterprises can reduce storage costs by improving and optimising how the available storage is shared among cameras. The traditional storage approach uses network video recorders (NVRs), which are PC servers that act as gateways to a directly attached RAID array or SAN storage. By moving the intelligence to the edge, it is possible to configure an IP camera or IP encoder to stream directly to an iSCSI RAID array, bypassing the NVR and the associated capital and ongoing costs.

The difference in this architecture is particularly noticeable in dispersed systems where the WAN lacks the bandwidth for centralised recording. Instead of installing NVRs in each building, which can be a maintenance nightmare, enterprises can install storage that is directly attached to IP cameras or to IP encoders.

Consider a university with 20 buildings and 25 cameras per building. Each camera is recording high-quality video at 2.8Mbit/s per camera for 90 days. With an NVR solution, the university would need 1,500TB of storage and 40 NVRs spread over each location. With direct-to-iSCSI recording, the 40 NVR PCs are not required.

Multiple cameras can share the direct-to-iSCSI RAIDs on a local recording network, keeping the recording load off the main local or wide-area network. Video traverses the backbone only when personnel need to review stored video. When this occurs, the video can be shared and searched by anyone with the proper network access.

This approach also means that enterprises do not lose recorded video due to network outages or NVR server hardware or software failure.

Video-recording management software also can act as a traffic cop — helping distribute video in 1GB blocks across various iSCSI disk arrays on the network. This allows for higher storage limits for cameras that are capturing video in sensitive or high-traffic areas where alarms are frequent.

Also, if one disk array fails, the recording-management software will simply redirect video to a backup device. That results in better disk utilisation, better load balancing and greater reliability. This architecture is also simpler, with fewer items that can fail, and less hardware and software for IT personnel to manage.

Among enterprises, there is no doubt that the use of video surveillance is growing, making it important that IT managers understand the methods to eliminate issues that can cause network congestion and failure. Recording video at the edge is a simple and common practice that can make IP video surveillance systems easier to manage and more cost-effective.

Bob Banerjee is the product marketing manager for IP video products at Bosch Security Systems. This article first appeared in Network World.