Telecom wholesaler Brightlink Communications routes 7 billion to 8 billion voice minutes a year through its Atlanta data center, connecting calls for mobile providers such as AT&T, Verizon and Sprint.
Service assurance for that many calls can be tricky, since those calls hit Brightlink's network monitoring tools like a fire hose. To cut down the amount of raw data that its monitoring tools had to process, around 18 months ago Brightlink deployed
By monitoring only SIP signaling packets, Brightlink has better insight into the health of the 12 Gbps stream of voice data that passes through its data center, according to Joe White, Brightlink's vice president of operations.
"We wanted to have full captures of every single call that was coming through our network. However, with the amount of data we push, if we were to mirror everything and throw it to a few [monitoring] boxes, it just couldn't be done. It would blow them out of the water," White said. "We tried to do that at the beginning, but the amount of RTP [real-time transport protocol] data was too much for normal network cards and Dell servers and [monitoring] software to handle."
SIP signaling packets tend to be a major indicator, and source, of IP voice problems, and yet they compose only a tiny percentage of the amount of data in voice traffic. White realized that a network packet broker would allow him to filter out RTP packets and send only the SIP signaling packets to his monitoring tools.
"We found that Net Optics would allow us to peel off the signaling portion and send every single bit of signaling to one of our monitors," White said. "At that point, we had full call ladders showing the back-and-forth of the calls as far as signaling goes. So we can see the different messages we're sending to various networks for a call flow."
Visibility into SIP signaling data allows Brightlink to determine exactly who and what caused a call to fail.
"With signaling we can see who sent what message to whom and what they replied with," White said. "We can look at call flow and say, 'We sent you some sort of signal and you never responded, and that's why the call dropped.' It helps better diagnose what happened on the call."
Brightlink can also use the Net Optics appliance to send full data streams -- both signaling and RTP packets -- to monitoring boxes when needed.
"We can drill down deeper on a case-by-case basis, and on certain premiere customers we could filter and say, 'For these subnets, I want to monitor all the voice quality and all the RTP.'"
The filtering capabilities of Net Optics have enabled Brightlink's network operations team to go from reacting to trouble tickets to solving problems before a ticket comes in.
"Before Net Optics, we knew how many calls were up, but we couldn't actually tell you the specifics of what happened in a call flow," White said. "Having that ability to look into the actual call flow allows us to respond a lot quicker, because we're not going to hunt down the call data record and ask our vendors what they think happened. We're able to diagnose exactly what happened. On top of that, by being able to sample media going through, we can proactively look for audio events -- packet loss, things that may be occurring on the network -- before we get a trouble ticket. So it not only allows us to respond quicker to tickets that come in and issues that customers report, but it eliminates things in the network before they actually become an issue."