I did quite a bit of research, and wanted to run this by the forum to seek some advice.
We have a client whom we setup an IP system for (moving from analog) of 55+ cameras of varying resolution from 1.3 to 3MP each. We've set recording as per our experience and his requirement, and some how neglected early on the issue of live display. Upon implementation, the live display for all 55+ cameras - which is a requirement of the client as active live monitoring exists - the maximum FPS we could get for the live display is 15 FPS, and the client want's his typical 24FPS minimum he was used to on analog.
We have the cameras spread over two separate NVRs based on with high specs running windows embedded OS.
I understand that this is a throughput issue, but the more I dig into it the more confusing it gets. I want to understand where the real bottleneck is, to explore options to fix it.
My conclusion on the matter is that it really is a NVR processor throughput issue, since all the load is on the processor (client requires continuous recording - not motion activated - for at least 10 hours a day). If this is the case, in theory adding more NVRs would solve the issue - not ideal, but at least a clear answer. It may be also more affordable in terms of storage vs adding iSCSI DAS storage.
One of the suggestions from the vendor is to use an additional piece of software, but it would require that the recording FPS match the required viewing FPS; which leads to crazy storage costs. They mentioned something about despite the camera being capable of generating multiple streams, there is one stream and it would drive this solution!
I am left with a couple of doubts over weather the issue could be a client PC spec issue since the live viewing is done on the client PC, or if there maybe a software hardcoded limitation on the throughput allocation for client live viewing ( I have been lead to understand that throughput is divided into recording, local live viewing, and remote viewing).
I know that some of the issues are very much vendor related, but I want to understand the logic that goes on here, and really identify where is the bottleneck.
Thanks
