Hi,
i want to set up a system with 6 AXIS 206 kameras and one AXIS 221.
Capture 32pfs from each camera or at least 16 in idle mode then 32 when alarm. Picture size would be 640 x 480. What kind of computer/server would i need for this?
Is it possible to get clear picture even when people are moving fast by the cameras? So that faces will be recognized?
Hardware needed for 8 kameras
I'd start with a Dual CPU server board and try with one quad core 3GHz CPU, If it turns out not to be enough You can populate the second socket with another quad core. Make sure that you have a couple independent PCI buses so definitely a server main board and use one of these to put an ether card ( So you can split up the the incoming/outgoing bandwidth but you may need to spit up the incoming traffic as well).
Might want to consider the Axis 207 in place of the 206.
Might want to consider the Axis 207 in place of the 206.
You will also need a good chunk of memory, so a 64bit OS will be required. I would also set up 1.5TB of storage or more. Now if you drop to 320x240 you will actually be loosing very little data as the horizontal lines are interlaced anyway. But you will save a lot in CPU and memory. Specifically a good core2duo on a p35 based desktop motherboard, 4 gig ram, and a 500gig hd would be plenty.
is graphics card important as well?
cordel, why should exa split outgoing/incoming traffic? Is it because 6(six Axis 206 cameras)*640x480(resolution; ~100kB max)*30(FPS)+1(Axis 211)*640x480(resolution; ~100kB max)*30(FPS)= 21MB/s and the usual speed of ethernet is 12.5Mbps? And what do you mean by outgoing traffic?
cordel, why should exa split outgoing/incoming traffic? Is it because 6(six Axis 206 cameras)*640x480(resolution; ~100kB max)*30(FPS)+1(Axis 211)*640x480(resolution; ~100kB max)*30(FPS)= 21MB/s and the usual speed of ethernet is 12.5Mbps? And what do you mean by outgoing traffic?
There is the camera traffic, and there is the users accessing the web page (or what I said as outgoing). You don't want your captures to suffer if badwidth gets tight from a bunch of users watching the monitors. Of coarse most all server main boards these days come with GB network so the addition of a GB switch would give you max potiential for expansion.
While the images on the network will account for ~100k (120k typicaly) for each image, there is also the overhead of the transport layer going through the rest of the levels of the OSI model so your useage will be a bit higher. Graphics card is not all that important unless you want to install a desktop, Using a browser on the same machine as the server though will suck up alot of resources and leaving a browser up usualy crashs the machine, avoid it if you can.
While the images on the network will account for ~100k (120k typicaly) for each image, there is also the overhead of the transport layer going through the rest of the levels of the OSI model so your useage will be a bit higher. Graphics card is not all that important unless you want to install a desktop, Using a browser on the same machine as the server though will suck up alot of resources and leaving a browser up usualy crashs the machine, avoid it if you can.
ok, thanks. However, if we say I have 8 cameras and they all are around my house and connected to the 10/100Mbps switch. And then I have a server, which has two 10/100Mbps network cards(both of those cards are connected to the switch). Both network cards in the server have unique IP address. First one has 192.168.1.34 and the second one has 192.168.1.35. Should I say to my cameras, that they should save their MJPEG stream to the 192.168.1.34 IP address and all users should watch the picture from 192.168.1.35? And it is probably smarter to bound the 192.168.1.35 address with internet as the traffic should probably be lower on that network card. Did you mean that by splitting up the incoming/outgoing bandwidth?cordel wrote:There is the camera traffic, and there is the users accessing the web page (or what I said as outgoing). You don't want your captures to suffer if badwidth gets tight from a bunch of users watching the monitors. Of coarse most all server main boards these days come with GB network so the addition of a GB switch would give you max potiential for expansion.
While the images on the network will account for ~100k (120k typicaly) for each image, there is also the overhead of the transport layer going through the rest of the levels of the OSI model so your useage will be a bit higher.
Actually I meant it would be wise to put the cameras on their own switch so that they are on their own broadcast domain, you want as little latency as you can have for the cameras, Hubs are bad because they create allot of collisions (ka collision domain) and switch's while they correct that still send broadcast traffic to all nodes attached. It is optimal to have the cameras on their own network, while for eight cams may not be a requirement in a residential setup.
cordel wrote:Actually I meant it would be wise to put the cameras on their own switch so that they are on their own broadcast domain, you want as little latency as you can have for the cameras, Hubs are bad because they create allot of collisions (ka collision domain) and switch's while they correct that still send broadcast traffic to all nodes attached. It is optimal to have the cameras on their own network, while for eight cams may not be a requirement in a residential setup.
What kind of broadcast data switches send to its nodes(IP cameras)? Does this broadcast data take significant amount of bandwidth? What I mean is that 10/100Mbps switch has "power" for 8(number of ports)*2(uplink/downlink)*100Mbps. And Cat5e has a max transfer speed ~150Mbps. I can see no bottle-neck here
Or is here a issue with latency, in other words cameras are not so responsible when they are attached to the large switch with other devices?It's not cameras traffic that I would be concerned about. It's little Jake getting on the computer and playing those java/flash based games and video, or if your playing with open-wrt/ddwrt and have a bug that does not behave correctly and suck the life out of your network (which happens often with users who do not reset there router after updating the firmware is a good example that happens often and they wonder why the network is slow and complain of the forum that the firmware sucks 
). I never said that the cameras would create broadcast storms, but other things can, and if you are not aware of and know how to deal with, then the easy thing to do is have the cameras on their own and you don't have to worry about cams going off line because of latency. The other part is your going to want to view your cameras, personally when I pull up my cams, I do so in a montage view at the same rate that the cameras are going so I can see them all, that uses bandwidth as well and you have to take into account that TCP has a bit of overhead so you got to count the envelopes as well. It takes may frames for one image at 640x480 to go across the network so in the end it could take you 140k to send that 100k file. I do not recall what the over head for TCP is so the 140k is not literal, some one verse in CCNA would be better versed at explaining all that or you could pick up a CCNA book if your interested.
You can basically set it up any way you like.
The first post said nothing about residential and that there would not be any expansion. You might be able to get a dual core system to cope with 240 fps but I have never put a IP cam system together with that high of a frame rate. And we should really not be hijacking someone elses thread.
). I never said that the cameras would create broadcast storms, but other things can, and if you are not aware of and know how to deal with, then the easy thing to do is have the cameras on their own and you don't have to worry about cams going off line because of latency. The other part is your going to want to view your cameras, personally when I pull up my cams, I do so in a montage view at the same rate that the cameras are going so I can see them all, that uses bandwidth as well and you have to take into account that TCP has a bit of overhead so you got to count the envelopes as well. It takes may frames for one image at 640x480 to go across the network so in the end it could take you 140k to send that 100k file. I do not recall what the over head for TCP is so the 140k is not literal, some one verse in CCNA would be better versed at explaining all that or you could pick up a CCNA book if your interested.You can basically set it up any way you like.
The first post said nothing about residential and that there would not be any expansion. You might be able to get a dual core system to cope with 240 fps but I have never put a IP cam system together with that high of a frame rate. And we should really not be hijacking someone elses thread.