While there is a little merit to this that if the network is lost, the camera can still record locally but the reality is the network is not typically the weak point in the system and on a LAN, a 99.999% up time is the target IT managers aim for as down time cost the company money.
Also if a LAN does go down, many cameras are using PoE to run the camera so the camera is off line anyway unless it has a separate power source and UPS, which most installations do not.
With wireless, providing they are correctly designed you should get similar uptime results (not talking domestic 802.11 links).
However the way video trickling is being marketed is leading end users to believe that if they have cameras at Site A, they can have the video automatically appear at Site B over night, allowing high resolution local review and off site archiving.
Scenario 1 – The LAN or WLAN going off line
Ok so hypothetically the LAN connection is lost for 1 hour and the camera has power to it. Using a CCTV design tool, I can calculate that a typical HD 720p IP camera will use 1GB of storage in that hour based on a 2.25Mbps bandwidth cap. This may be higher for some cameras and scenes.
Your 100baseT LAN has a perceived 100Mbps throughput but the actual maximum is 76Mbps. Then we reduce that by the handshaking overheads of around 8% so 70Mbps. But the typical network chipset in the camera is only able to transfer 32Mbps.
To move 1GB of data over this AND the camera’s normal streaming of 2.25Mbps will consume around 5 minutes and you may find the live streaming/recording does not behave well as you are maxing out the chipset. That is in a simple overview a usable feature when you ignore the other factors that may impact on the real world results.
Scenario 2 – Download the events only to an offsite storage
If we now look at a system that is using some form of event tagging and we get 10 events per day, each recording for 10 minutes. Using the same figures as above but over an ADSL connection of 1.5Mbps (T1) as people believe they can, we soon see that it is not viable.
Each event will consume 166MB of storage so 10 will be 1.66GB we need to transfer each night.
On a perfect WAN, with no retries and a TCP overhead of 15% (1.91GB) will take around 3 hours to download. That is for 1 camera only. So a small system of 5 cameras will take 15 hours, which means it will not be in off peak times and the data plan to move 5x 1.91GB by 22 business days a month is a whopping 210GBnon centralised per month plan.
From my view point, there is not really a market for this technology at this stage and even the VMS manufacturers do not have it working 100% as documented in an IP Video Market article. If your network is designed well and you use a no centralised server based architecture, the reality is you are not likely to lose any video.
Most products can download events to a remote location so the more important factor is how much storage each event consumed and what event management does it offer so you can determine the bandwidth requirements.
