Making the most of IP
Despite the high number of IP cameras Swinburne uses — and high definition ones at that — Goetze says managing network traffic has not been an issue.
“The amount of data is not as much as people seem to expect, especially with the new video codecs you can use [like] H.264,” he says. “One hour of video on a high definition camera if you are running motion JPEG can be about 1GB of traffic, but if you are running a different compression it can be less.
“We are getting 43 kilobits per second (Kbps) of data streaming from each camera. It is not a huge amount but it adds up when you are running [cameras] in the hundreds.” Goetze says to get around the data challenge Swinburne put a recording server in each of its campuses eliminating the need to pull data from the cameras across its backbone. Further, if the network ever loses power, each campus can continue operating its IP cameras under their own steam.
Rather than opt for centralised storage or leverage off its existing SAN infrastructure the university opted for 12TB of direct attached storage for each recording server. On average each server handles 50 cameras, giving the university four to six weeks’ worth of storage capacity of video footage.
“After four to six weeks we just roll over and that satisfies our needs,” Goetze says. “We keep with what the Australian recommended standards [for security feed footage storage] are, which is about four weeks.”
The MCC’s Liddle says one of the biggest benefits of a move to IP-based cameras has been the ability to consolidate six racks’ worth of surveillance gear down to half a rack – thus minimising its data centre footprint – and increased flexibility and control of its surveillance system.
“It’s basically an IP IT system with devices hanging off the end of the network," Liddle says. “Working with our security guys has been good and we have been able to leverage off each other’s expertise. That flexibility in the IP model to be able to chop and change cameras on the fly has been really good."
Another benefit had been the ability for multiple users to access the system from different locations within the MCG grounds increasing the ease with which staff can use the system.
The MCC opted to run its IP cameras over copper with a fibre ring running around the MCG grounds for the backbone and power the cameras using Power over Ethernet (PoE).
“You are probably looking at a third of your cost for the cabling and two thirds, depending on the quality, for your cameras,” he advises. “It’d definitely a large investment in your infrastructure for your backbone.”
For storage, the MCC uses its existing IBM-based storage but has segregated its IP camera storage due to need to constantly write data to disk from the cameras. Footage is kept on hand for 15 days before being recycled, meeting both storage and privacy requirements.
To address the concern of using IP cameras in a public environment — and particularly a people counting application — Swinburne University of Technology steered away from facial recognition and instead applied live analytics that guarantees individual subjects aren't personally identified.
“Essentially all it stores is counts so it knows six people came in and five people went out, but it has no idea who they are,” he says.
Then again, Goetze hasn't seen a push back around privacy and, in fact, the university’s staff have actually requested the more cameras to be installed for their own safety.
“That was our three main goals: property protection, the health and safety of staff and the students, and to ensure that security was designed from the start [with privacy in mind],” he says. “We put all the cameras on their own separate VLAN which is locked down — only the servers can access those cameras and only security officers can access those servers with a log in, which lets us track what they are viewing.”
Follow Tim Lohman on Twitter: @TLohman
Follow Computerworld Australia on Twitter: @ComputerworldAU