Update: C.H.I.P has a $9 computer available (for pre-order), this brings us even lower. For a very small form factor, there are now a host of micro-computers out there (where micro refers to the physical size).
You can get a real cheap DVR for around $100 bucks (without storage), that usually support 4 analog cameras, but we’re aiming a bit higher here. We’re going for something that supports IP cameras at various resolutions and we should be able to support more than just 4 cameras for $100, preferably 10 or 20 cameras per node.
With the HW platform coming in at $50, that leaves me $50 for a drive. A 1 TB external drive can be had for $55 on Newegg.com, so I’m on budget.
If I allocate 1 MBit per stream then 20 cameras would be 20 Mbit. The HW has a total IO bandwidth of 450 MBit, so doing 20 cameras should be cakewalk. Even if we go to 2 MBit, which might be needed for decent quality.
If my calculations are correct, then I should be able to store 9 days of continuous recording from all 10 cameras. If that is not enough, larger capacity drives can be had for $100, or I can mount a NAS drive (storage is cheap).
I decided to go for external storage rather than directly attached as it offered a pretty neat feature. If something happens, and I need time to review all the recorded video from all the cameras, I can simply plug in a different drive, and keep recording on that. If the drive is physically fixed inside the enclosure, then you have to copy the drive to an external drive, which can be a long and painful process. In a sense, I am treating the drive as a large VHS tape. This is not the perfect solution for all applications, but it doesn’t have to be. It just has to be perfect for the application that I have in mind.
Can this low-cost HW do server side motion detection? The answer is a resounding NO. But setting up motion detection on the edge is often trivial, and it allows for much better scalability than doing everything on the server. On a side note : If you ARE going to do everything on the server, then please, don’t buy cameras that have very advanced edge capabilities, just get some cheap ones with a good imager and optics.
Get some PIR detectors, and hook them up, then turn on VISIBLE light when they trigger in the armed state. For example; you don’t want the floodlights to turn on when a deer walks across your lawn, and you are at home. So when the system is not armed the PIRs don’t turn the lights on. Visible light is a much better deterrent than IR which a) provides b/w images with false nuances, b) does not alert the neighbors and c) must be on at all times for the camera to be able to do motion detection (waste of energy).
At the bare minimum, I can set all of this up for $100 (with the lousiest interface you’ve ever seen, and an install process that takes 3 hours of downloading, compiling, configuring, sudo’ing etc.), but it CAN be done.
The “box” can also do cloud recording, but I have decided against that. There are 2 primary reasons for this : 1) my internet is just too slow and unstable and 2) I am not thrilled with the idea of my video residing at some random server with google or amazon.
This suggests to me that unless you (at the current retail cost of VMS appliances) are not making head over fist, then you are in trouble. Granted, the appliance is not appropriate for all installation types, but there are many places where an appliance is easier, safer, faster and cheaper to install. If a n00b like me can bang together an “appliance” over a weekend then just think what can be done if a few of the guys got together and turned this into a product?
(oh, and as an added bonus, I can stream from my phone to the recorder too)