When the aviation services arm of CSC began migrating its world wide customer base to converged MPLS networks, it hit a snag - traffic visibility, or rather invisibility. In the past, dedicated networks had made it relatively simple to track who was using what, but the converged nature of MPLS meant that one pipe carried many different traffic types.
The problem is that the packet labelling used by MPLS provides no visibility into application flows and activity, explains CSC network engineer Tore Enger. It also makes it hard to ensure that applications are getting the quality of service (QoS) tags, he says, adding that while CSC's various telecomms service providers could supply the necessary reports, they were neither timely nor in the format CSC's staff preferred.
"WANs report on usage but not applications, volumes, percentages, etc," Enger adds. "It would cost them more to provide that, and not all carriers would report at the same time using the same tool. The biggest problem is when a user calls in and complains of a slow system, and the reports can only tell us utilisation on 15 or 30 minute intervals."
From its base in Norway, CSC provides network services to around 20,000 users at 350 customer sites in the airline industry, some 90 of which have so far been moved over to MPLS. It had previously used SNMP and NetScout's nGenius performance management software to control its ATM and Frame Relay application traffic, restricting unknown traffic to five percent of total volume for example.
"With the NetScout system we can get finer intervals and see what the link is used for and who it using it to talk to who. It means that we can quickly see who's using the large part of the bandwidth, for streaming radio or video, say," Enger explains. "We use a packet analyser with a stategically placed probe, and sniffers to analyse traces."
The big problem was the advent of MPLS meant that this method no longer worked on those lines. One option would have been to install a probe at each customer site, but this would have added considerable cost as well as administration work. "We used a bit of SNMP network management, but we had to travel a lot and use sniffers," Enger says.
Bringing MPLS into line
Fortunately it was around this time that NetScout introduced a Site Monitoring and Reporting add-on for nGenius, as a free upgrade for existing customers. CSC picked this up and found that it solved the problem by identifying each remote site coming into the router and where the traffic is destined, validating QoS priorities, and so on.
"The site monitoring feature means we can have the same reports for MPLS, ATM and Frame Relay, so if the provider changes the pipe it's no problem," Enger says.
He adds that while MPLS gives prioritisation, you still need to report on that. For example, a misconfigured router could lead to almost everything getting the highest priority, and not just the voice over IP traffic that demands it, or a CSC customer could install a new application that affects the network.
"There's also the fact that a customer problem might only happen one or two times a week," Enger says. "With this product we can analyse immediately, instead of hunting through a huge report later. We have set limits on response time too, alerting the administrator with user and application information for that link. So we can see a slow link and analyse it before the customer complains."