Risks of Overloading Video Surveillance SystemsBy John Honovich, Published on Mar 29, 2009
Cameras, software and recorders are all becoming more powerful and complex. Higher resolution, multiple streams, multiple analytics and more powerful CODECs greatly enhance the potential of video surveillance. At the same time, using multiple features together can routinely overload a video surveillance system, creating significant problems that are hard to resolve.
Traditional Video Surveillance - Simple but Limited
In the old days, you had analog cameras and DVRs - both of which were very limited in what they could do. Because they were so limited, it was hard to misconfigure or overload the system. A recorder only had so many inputs and an analog camera could only send a single fixed stream.
The Claims of Today's Video Surveillance
Today's video management systems and cameras can use H.264, can handle multiple megapixel streams and run on-board analytics on video. It is becoming very common for companies to market all of these advantages.
The Problem for Users
The problem is that it is almost impossible to run all, or even many, of these functionalities at the same time. Making the problem worse, this is generally only disclosed in the fine print or footnotes of the marketing material.
Users and integrators need to be careful about designing systems with these products. 'Optimistic' designs based primarily on vendor marketing can deliver systems that significantly under-deliver or break down in real world use.
Recording both IP and analog cameras is a key claim for hybrid DVRs. The critical question is: how many IP cameras and at what resolutions and CODECs can this be done? While a hybrid DVR may record 12 IP cameras at MPEG-4 with CIF resolution, it may not be able to record even a single 3MP camera using H.264.
A similar concern exists with video analytics. Running video analytics inside of the DVR offers a lot of potential. The challenge is knowing how many channels of video analytics and what type of video analytics can be run.
The more video analytics you use, the less IP cameras you may be able to support. Plus, the more activity on your cameras, the less you may be able to manage as well.
While there are clear benefits for using hybrid DVRs, there are risks as well.
IP Camera - Example
IP cameras are more frequently being marketed as platforms that can send multiple streams, use multiple CODECs and run a variety of video analytics.
Just like with hybrid DVRs, the problem is you cannot do all of these at once.
More critically, a growing belief exists that any IP camera can be a smart camera. Many IP cameras can have some smarts but almost all IP cameras on the market are quite limited in how 'smart' they can be. Agent VI [link no longer available] is one popular approach of taking existing IP cameras with moderate "horsepower" and adding video analytics. However, Agent VI still requires additional server side processing. This does not make it bad. You simply need to be aware of this in your designs. Via:sys [link no longer available] does claim to run solely on existing Axis and IQinVision cameras but you should be extremely careful about its level of false alerts and how it can perform in adverse environments.
IP Video Surveillance Software - Example
One of the most attractive claims of using IP Video software is that you can greatly expand the number of cameras you can record on a given server. Unlike cameras and hybrid DVRs, you also can choose your own hardware to optimize it for a given load.
Even with this flexibility, you need to be careful about the settings of cameras used. While vendors often talk about recording hundreds of cameras per server, they usually mean standard definition cameras using MPEG-4 and only motion detection. Once you start using multi-megapixel cameras, H.264 CODECs and video analytics, the number of cameras per server can drop dramatically.
Why does this problem exist?
Manufacturers are careful to keep costs as low as possible to remain competitive. For instance, IP cameras and hybrid DVRs could support far greater loads if the manufacturers used faster processors, greater memory, etc. The trade-off is that this costs more money and would force the manufacturer to increase the price of the product significantly. Most manufacturers have decided that the best route is to keep costs low at the expense of greater complexity and risk of overloading. (I agree with this.)
Over time, as computing resources become cheaper relative to the growth in video surveillance technologies, manufacturers will no longer face this trade-off and the problem will be resolved. However, for the next few years, given the explosion in resolution and the demands of video analytics, system designs must carefully factor in this trade-off.
What symptoms or issues will occur?
The risks are similar to that of over-loading your PC with too many applications. When you have too much running, the PC can become much less responsive, if you are watching a video, the video may stutter and eventually your PC may lock up. Since cameras and recorders are computers, it's not surprising that they are impacted in a similar ways as to PCs.
The specific video surveillance risks are:
- Frames may drop. Even if you set the frame rate for 15 frames per second, the camera may only be able to send 8 frames or the recorder may only be able to record 7 frames, etc.
- Video analytics may miss events (false negatives). Dropped frames may cause the video analytics not to catch an event.
- The recorder or camera may become very slow or unresponsive. If you do a search or request video, the system may not be able to deliver as it is blocked by other processes already running.
- The system may crash or need to be manually rebooted.
- Resolution of each camera
- Frame rate of each camera
- CODECs of each camera
- % Motion activity on each camera
- Number of channels of video analytics being run
- For cameras, how many streams being used