System Requirements of Hypothetical System

[scottclayton]

Hi There, Ryan, et. al.,

I'm desiring to build a new surveillance system for a store and mall, and I have determined that about 57 cameras running at 720p—or a mixture of 720p and VGA—would be about perfect. Unfortunately, I haven't been able to determine from the help-documents, or elsewhere on the internet, what CPU and memory loads would add up to, and what kind of system would fit well. It would be nice to have it all running on one computer, but if that's significantly more than two computers, then I would sooner build two computers. AMD makes an affordable 16-Core 2.1 GHz processor, and I could stick two of those in a single machine, and pop in maybe 64 GB of RAM, 4 * 3TB HDDs in RAID-5, and maybe 5 * USB 2.0 expansion cards, since the cameras would be USB webcams, and then I could keep it down to two USB webcams per USB channel. I'm just not sure if it's overkill, or underkill, or on-par-kill. Our current setup is using analog cameras converted to digital, with a resolution of QVGA, and 29 of them run on an old Pentium 4 with 128 MB of RAM and two 300 GB HDDs in RAID-1, so my benchmark isn't exactly Apples-to-Apples, or even Carcassonne. I'm not sure how Vitamin D stacks up against competitors with respect to processor/memory requirements. Our current system also uses motion detection, and, although less sophisticated, it's not burdensome for the operator when false-positives are generated—literally taking up an extra half-second of time, once every few days.

Anyway… happy Monday!

[ryan]

Hiya Scott,

I'm happy to get into further detail, but here's a quick overview:

CPU is much more important than memory. I think that's going to be one limiting factor in your setup. Though CPUs are faster than when we wrote this, it is still relevant more or less:
http://www.vitamindinc.com/support/ref_ ... ps_01.html

One reason it continues to be relevant is that even with faster CPUs other aspects tend to become the limiting factor. As with any large scale system like you are proposing you're going to have to become a great sysadmin and understand every part of the system. It's great that you mention raid, because disk access is quickly going to become a factor. Even 3 years ago we had a customer running 10-12 VGA cams I believe on a single system, but they had to move to a raid system as that was the limiting factor. Even then you'll need to get into things such as hardware vs software raid, raid types, disk speeds, performance, etc...

Definitely hard to compare to other video systems as you note because VDV does SO much more computation on each stream vs just decoding and encoding video. Your current setup isn't really going to be a benchmark for a VDV system. Even general internal system IO is much higher given VDV's extensive tracking and database operations for example.

With regard to using USB webcams, I think you might run into trouble monitoring a mall like this. The official spec for USB cables isn't too terrible far (at least on building scales) as I understand. Might also start needing to worry about the specific specs of your usb hosts, your computer's bus, etc..

Hope that helps, feel free to ask further questions and I'll do my best to answer!
- ryan

[Semper Vaporo]

I use an "Active Extension Cable" for my USB webcams from StarTech (dot com) that is 16ft long. The packaging claims you can cascade up to 5 of these cables to reach 80 feet (25 meters). I have found that I can get another 5 ft more using a plain extension cable on the end of 2 of these cables in series. I have not tried the claimed maximum distance with extra non-powered cables so I don't know if one could go further; I suppose one might go further by using short plain extension cables between each of the Active cables, but it might be a hit or miss proposition as to reliability.

Each of the cables draws current from the computer's USB port. The USB spec says each device can draw up to 500 mA and the spec on the cable package says each amp draws 5 mA, so that does not seem to be a limiting factor unless the camera on the end is drawing close to 475 mA.

There may be other cables on the market that have external power that might be able to go further or have more in series to gain greater distance.

There are also products that can convert the USB signal to wireless (WiFi or otherwise), or to use the building's power wiring for signal purposes, with a receiver at the PC to revert the signal to USB for the PC.

Another consideration is the number of video signals that a USB port can handle. I am not entirely certain at how this all works but it seems that in the PC there are circuits that connect several USB ports to the main data bus. You plug a device into a socket where there are hardware drivers that supply the power to the device and transfer signals between that socket and the "Host Controller". The Host Controller is what connects multiple drivers to the Main Data Buss in the computer. There is a limit as to how many drivers a Host Controller can handle, and I assume there is going to be some limit as to how many Host Controllers can be on the Main Data Buss.

There is also a limit as to how many of these devices can be configured to pass data at Video speeds.

My experiments with trying to get multiple USB webcams to work indicate that at least one port on each Host Controller will not be assigned permanently to a Video device (I think that is an operating system function). I think my dedicated Surveillance PC has 4 ports on each Host Controller and I can plug in 3 cameras into each and get them to work just fine (with certain limitations to be discussed later), but when I plug in one more camera, it is recognised, but will not work until I unplug or otherwise disable one of the other cameras. And the computer manufacturer didn't put all the sockets in groups per each Host Controller which complicated the process of discovering why one camera stopped working after I did some redressing of the cables (I accidently plugged one camera into a different socket than before I started re-dressing the cabling... or maybe I should say, I accidentally had plugged the cameras in so as to have them distributed in a good way to begin with and screwed it up in the clean up!)

Another limitation is that I have had a very difficult time getting two cameras of the same brand or model to work at the same time. I was buying cheap Webcams and was using various brands and not having much trouble with them, other than they were cheap cameras that had poor low light capabilities and tended to smear motion. So I found some better cameras on a clearance table and I bought two of them. Either one would work fine by itself, but when I plugged in the 2nd one, the software driver did not distinguish one from the other and no matter what programs I ran to view them, the image displayed for each window would periodically switch to the other camera! (Kind of weird to watch them flop back and forth showing different images every few seconds.) And not too long after that I'd get a BSOD ("Blue Screen Of Death") and have to reboot the computer.

I think that is a problem with the software drivers not being written to "enumerate" multiple identical devices. I am sure it could be fixed in the drivers, but it is a matter of convincing the programmers to do so.

I have 3 different models of Microsoft camera and none of them will work if any of the others are installed. I have found two different Logitec model cameras that will work together, but I also have some "similar" Logitec models and they will not work together. I have some "rebranded" cameras (made by Logitec, but under a different name) that use the same drivers and they are mutually exclusive with the other models from Logitec or other rebranded ones.

I am presently using: Microsoft HD-5000, HP-3100, Logitec Pro-9000, and a Logitec Webcam 120 and they are all happy with each other as long as I don't plug them all into the same Host Controller.