How storage virtualization bridges gap between budget and needs

A comparison of data storage options for video surveillance


A recent report by IMS Research found the video surveillance storage market is roughly $1.2 billion. This market is growing at more than 60 percent annually based on the facts that more cameras are being installed, higher resolution cameras are in demand and longer retention times are required. The recently thwarted New York City bombing demonstrates the predominant growth of surveillance - 58 private camera views were available following the incident. And this number is far less than the number of views that would potentially be available in London, arguably the city with the most surveillance cameras deployed globally.

As the amount of video data grows, the viability of traditional DVRs and NVR with local storage is breaking down. With these systems, a fixed amount of capacity is available for video streams that may vary in needs over time. Determining what capacity to buy is a challenging task since the amount of storage depends on an estimate of resolution, frame rate, compression and motion detection by camera type. Therefore, the nature of the DVR or NVR infrastructure generally leads resellers to over-provision capacity for each system by up to 50 percent since fixed capacity is not easily expanded and has relatively low reliability.

More advanced compression technologies such as H.264 help bring storage requirements down when there is minimum motion in the field of view. However, the variable nature of the bit rate exacerbates the capacity estimation problem since requirements can be off by as much as 4x if motion is higher than expected. The market will continue to see improvements in compression standards, but the growth in the number of cameras, resolution and retention times will continue to drive the need for scalable and highly reliable storage solutions.

Effective ways to increase storage beyond a DVR or NVR
A storage area network, or SAN, is ideally suited for a centralized site such as an airport, casino or prison where there is plenty of bandwidth on the local area network. In a SAN environment, video recorders or archivers maintain their own file system, such as Windows, and these appliances talk to the shared SAN storage as if it were a local disk. Many archivers can share the storage and the SAN platform introduces more reliability over NVR/DVR systems because there is no single point of failure.

A network attached storage system offers shared storage for many archivers with an important difference - there is a file system on the NAS platform as well as on the local archiver. The advantage of the file system on the NAS platform is that it is easier to support a hybrid storage case as some storage occurs locally on self-contained NVRs/DVRs and extended storage is sent to a specific file on the NAS. Both SAN and NAS offer highly available shared storage, but the performance limitations of NAS file systems preclude their use for high camera count, high bandwidth installations. As a result, NAS products are typically used in lightly loaded distributed environments or when data is being sent across a WAN connection.

If shared storage is not required, then it is possible to simply add storage to a self-contained DVR/NVR with an external connector, which is normally fiber channel or serial attached SCSI (SAS). This is called direct attached storage because the capacity is tied directly to one server and cannot be shared. External DAS is best used when capacity requirements are well known, will not expand over time and where the performance requirements are fixed. For growing environments that might, for example, replace analog cameras with megapixel cameras over time, DAS offers little investment protection since a system cannot be expanded without a "forklift upgrade".

The beauty of SAN or NAS solutions is that these options don't require exact calculations since storage can be shared and scaled as needed. It is not easy to predict the amount of data generated by an individual camera because bit rates are dependent on assumptions on how much motion will be detected. The amount of motion can dramatically affect the amount of data needs and what the configuration should be.

This content continues onto the next page...