When it comes to video analytics, security operations have turned a corner. Gone are the less-than-reliable analytics of yesteryear. Video analytics today deliver more precise, informative, and far-reaching insights, to a degree that many organizations have stopped questioning whether to use video analytics, and are working to determine how and where to implement them to get the most value.
With this new focus, more organizations are weighing the differences and advantages of edge-based and server-based analytics, the two main options. But whether they choose to run the software on an IP camera at the edge of the network or to install analytics software on a centralized server to which all cameras would feed depends on their environment, existing infrastructure, and needs.
Here are 10 factors to explore with organizations selecting a video analytics deployment strategy.
1) Video analytics options
Determine what applications the organization will need to run on its cameras. While organizations can run most types of video analytics applications on the edge, they’re typically restricted to running one application per camera due to their limited processing power. Server-based video analytics are unrestricted by processing power, which means organizations can run multiple applications simultaneously and on any camera.
In remote or isolated locations, such as a substation, cell tower, or train depot, deploying excessive hardware, especially servers, doesn’t make sense, as these locations often aren’t easily accessible to maintenance crews. By placing the video analytics on the edge, however, organizations can rely on the self-sufficiency of the camera, which can do everything, including recording to its embedded storage. In this case, even if the network failed, the analytics on the camera would still run as normal. Server-based analytics in this kind of environment aren’t self-sufficient simply because they wouldn’t function independently in the case of a failure or mishap. And given the typical characteristics of remote, isolated locations, that’s bound to happen.
Does the organization care more about processing power or video quality? Obviously anIP camera on the edge has less processing power than a server, but sometimes it’s not enough to handle video analytics applications. Note that this can be compensated for with an on-board processor. On the other hand, cameras have access to raw, uncompressed, high-quality video, which in turn enables high reliability. Reliability only becomes an issue if the camera’s embedded processor is not powerful enough. Processing power forservers is a non-issue; they have plenty of it. Organizations can even reduce overload by defining a maximum number of cameras per server. Server-based solutions are very reliable. The only issue is the potential for poor-quality incoming video streams.
Again, this depends on an organization’s infrastructure. With low-bandwidth connections, analytics on the edge are ideal. Analytics take place within the camera, and when relevant, it sends an alarm to the center. Bandwidth isn’t much of a concern. With server-based analytics, the incoming video stream needs to reach a certain quality level in order for the video analytics to be able to effectively process the captured video. As quality is determined by frame rate and resolution, bandwidth is crucial.
5) Substitution capability
A server-based model offers more flexibility to substitute or switch out the analytics as performance needs shift. Organizations can just replace the software on the server with more effective analytics. When analytics are embedded within the camera, however, the process is more problematic. Most manufacturers use one type of analytics across their portfolio, so changing analytics software most likely means replacing every camera with a new brand.
6) Camera choice
In a server-based model, it doesn’t matter what kind of camera is being used, as long as the video quality is at the required level. Even existing analog cameras might work, as the analytics are applied at the center to the already-captured video feed. This is not the case with edge-based analytics. The higher cost of producing analytics-enabled cameras and the lower demand for them means fewer are manufactured and organizations have more limited camera choices.
7) Ease of installation
As mentioned above,camera manufacturers typically use the same analytics software across their portfolio. While it can be a bear to substitute, it does ease installation if all the cameras are the same, since the configuration of one camera can be replicated for every camera. With multiple camera types and models, however, installers will need to know how to configure each of them as they all use different types of analytics software. This is not the case with server-based analytics, which are camera agnostic and can be used with any camera or encoder brand and model.
8) Carbon footprint
When analytics are on the edge, the carbon footprint is minimal—IP cameras only consume a few watts of power and potentially only slightly more to run video analytics. With a server-based system, there are a few variables that contribute to the size of its carbon footprint. If the same server is being used for recording as well as analytics, the increase is negligible. When a dedicated server is used for analytics, however, it can use hundreds of watts of power, distributed across all of the cameras it serves.
9) Total cost of ownership (TCO)
When analytics are placed on the edge, the TCO is almost zero when taking into account factors like power, cooling, and rack space. An IP camera with an embedded SD card doesn’t take up rack space, uses minimal power consumption, and doesn’t require cooling from an HVAC system. Server-based analytics aren’t as clear-cut.If analytics are running on a Network Video Recorder (NVR), the TCO is negligible, as no additional power is consumed. Of course, if the analytics are running on a separate dedicated server, the TCO can be significant when taking into account all the factors required to maintain the server—rack space, power, cooling, and so on.
Camera manufacturers often include embedded analytics at a fairly low price point, but their affordability must be weighed against other factors. The increased reliability, power, and number of feature sets that organizations gain through server-based analytics also means that the cost per camera is typically higher.
These aren’t the only criteria that should be considered, of course. Licensing, for example, is also key. With edge-based analytics, licensing is assigned to a particular camera and in most cases can’t be switched to a different camera. With server-based analytics, licensing is assigned to a number of cameras, which can be more easily reassigned. Storage is another consideration. Since video analytics rely on captured video, storage is particularly important, especially when advanced forms of video analytics are being deployed.
Advanced video analytics require even more consideration
Unlike simple video analytics applications where the video is analyzed and checked for anomalies, then generates an alarm when applicable, advanced video analytics go a step further and also record metadata for future searching. For example, with advanced analytics, organizations can review recorded video and query the application to identify any number of criteria—shirt color, size, object type, speed, trajectory, aspect ratio, or even license plate numbers.
Leveraging the metadata generated by advanced video analytics requires that the video stream and its metadata are stored on the NVR. The video management system (VMS) must be able to interface with the NVR in order to query this metadata as part of a forensic search. As a result, video analytics and VMS vendors need to partner closely for seamless integration of the two systems. Because of this, it is not uncommon for highly sophisticated video analytics applications to be offered as a single source solution.
Carefully weighing the options
It’s clear that when a security operation is considering deploying either edge- or server-based video analytics, the environment in question will be a major determining factor. Remote locations with limited bandwidth are ideally suited for edge analytics, which can be an effective, affordable solution. Server-based analytics give organizations more options as they have more CPU power, can run more than one analytics application per camera, and are camera agnostic, offering freedom of choice and flexibility.
When you work with organizations to consider all of the options listed above, combined with the characteristics of the particular environment, the best choice for the organization should emerge.