With the increasing numbers of Megapixel/HD cameras, designing an efficient recording video surveillance system is crucial. There are a lot of components that can affect the video quality: cameras, network, storage systems and CPUs.
When it comes to choosing servers for VMS applications, ensuring an adequate server CPU and memory for both recording and viewing videos is very important.
When the server is under high CPU demand or the viewing station has not got sufficient processing power, the entire system will become less responsive.
The decompression of HD images from codecs such as H.264 or mjpeg in order to view them is processor intensive, and requires powerful machines to deliver high resolution, high frame rate images to the monitor. H.264 codec based systems encode a video stream to use less network space but then CPU power is needed to decode the stream so it can be recorded. The stream is then re-coded in H.264 and sent to the client machine where it is decoded again for viewing. Using camera multi-streaming you can set a view stream at a lower resolution than the recording stream but you still need to consider how many pixels are being displayed and at what frame rate to get the right performance.
Demands to view 16 or 32 cameras at HD quality, 25 FPS on a single monitor are unrealistic based on the current limitations of hardware image processing. As a rule of thumb, we recommend using a Core i7 processor with a maximum 100 fps at 720p per i7 CPU, that’s about viewing 4 standard HD cameras at full-frame rate, using H.264! You might get better based on a number of factors, such as scene, lighting and motion. This is why we use powerful Xeon processors in our servers, they aren’t cheap but provide greater power which means more cameras.
Tip: to improve the CPU performance and help the PC to work smoothly again avoid opening multiple cameras in multiple tabs. In fact, even if the tab is in the background, out of sight and out of mind, the PC still has to decode the H.264 streams open in that tab which will continue to draw CPU processing power.