Diameter: Managing the surge
At the Broadband World Forum event in Amsterdam in October, the same observation was made by two of the keynote speakers. Eelco Blok, CEO at Dutch incumbent operator KPN, and the European Commission’s VP for the Digital Agenda Neelie Kroes both warned that Europe is being left behind in high speed broadband, mobile and fixed, because the region’s operators lack scale.
Both speakers looked to the US and China as examples of markets where advanced broadband technologies are being deployed at pace, and both noted the benefits to the operators funding those deployments of huge scale in terms of addressable market. In Europe, Blok pointed out, there are around 150 mobile operators competing for a subscriber base of less than 750 million. This unfavourable ratio, he said, explained why so few European LTE deployments so far are “meaningful”.
Scale is a word we’re hearing a lot in relation to LTE; economic scale is needed to fund network deployment and the network itself must scale to cope with the soaring demand for mobile data. We’re less used to hearing about the importance of scalability in the management of Diameter signalling traffic— but according to Doug Suriano, chief technical officer at Tekelec, it is an issue crucial to the success of LTE deployments.
Tekelec recently published its LTE Diameter Signalling Index, which it believes is the industry’s first attempt to forecast the growth of Diameter signalling traffic in LTE networks. The firm expects Diameter messages per second (MPS) to grow at a CAGR of 252 per cent between 2011 and 2016, hitting almost 47 million MPS by the end of the forecast period. This growth rate, the firm says, is three times faster than the growth that mobile data traffic is expected to show over the same period.
As operators distance themselves from unlimited data plans and look to introduce true QoS distinction in their offerings, as well as contextual upgrades like application- or time-limited throughput boosts, the volume of signalling traffic is surging, Suriano says. Factor in the reality that the number of concurrent sessions per user is on the increase—with subscribers expected to become habitual multi-taskers across voice, video and browsing functions—and the scale of the burden starts to come into focus.
Policy and Online Charging functions will be the biggest contributors to Diameter signalling —“you won’t set up an LTE session without applying some sort of policy to it,” says Suriano. But more generally, the call flows in LTE involve a great leap in communication between different service elements, be they policy and charging systems, subscriber databases or devices and applications themselves.
Among the applications, VoLTE and video (especially with QoS) will prove the most demanding in terms of Diameter signalling, and operators are looking for ways to accommodate popular applications like Apple’s Facetime video chat service on cellular networks. This requires a ”robust architecture to cope with signalling and data traffic surges,” Suriano suggests.
Compounding the problem is mobility itself. Policies remain active for long periods of time; hours or days, he points out. Session state has to be maintained as the subscriber moves around the network in order to support the policies being applied to the subscriber’s usage.
The problem with Diameter, Suriano says, is that standardisation work at the IETF is still in the relatively early stages. Yet operators can ill-afford to postpone deployment as they wait for maturity, he adds, as these functions are the basis of mobile data traffic monetisation.
“The early pioneers didn’t see the need to have a core Diameter signalling infrastructure at the beginning of their LTE deployments,” Suriano explains. The expectation among operators and equipment vendors in the first wave of LTE deployment, he says, was that a separate Diameter control plane would not be needed until LTE networks were hitting ten or 20 million subscribers. “The common view was that a meshed architecture—the direct connections between each network element—could scale sufficiently. But they ran into scalability problems. The world went through the same thing with SS7 in the early stages, as well.”
“We started to see signalling surges on networks with fewer subscribers. This was caused by the shift from unlimited to usage-based data plans which require more signalling communications as well as the usual growing pains when new software is introduced into complex networks.”
It is the lot of the pioneer to trade painful learning against first-mover advantage.
“SS7 had the ability to handle congestion, handle traffic overload and throttle traffic built into the protocol,” Suriano says. “Diameter, on the other hand, was designed to be a more extensible protocol and it lacks that kind of sophistication. The idea was that the IP network would handle a lot of these jobs on a ‘best-effort’ basis. Now everybody is realising that we have to make the Diameter protocol deterministic, like we did with SS7, because you cannot handle these kind of problems at the IP layer.”
So how long does the industry have to wait for Diameter to mature? Suriano estimates that the technology is at the end of year two of a five-year development cycle. “The IETF has responded to this requirement and is leading standardisation to take Diameter to the next level of maturity where these kind of capabilities will be incorporated into the protocol.”
But, he warns, it will take perhaps another six months for the IETF to get the specification done for the next version of Diameter and then maybe another year to get that specification implemented into the equipment vendors’ product portfolios. Beyond that it may be another twelve months at least before the operators begin to deploy the new kit. So while the technology will stabilise in the next three years, he says, the industry as a whole will take a little longer to catch up. He reiterates his warning that operators should not wait for that process to happen before moving to a centralized Diameter signalling architecture.
“You have to start at version one to get to version two. If you wait for version two, you’re going to miss out on the LTE opportunity,” he says. “Operators need to work with vendors that are ahead of the game in building that functionality into their products, in advance of the standards body work.”
Tekelec is working with many of the leading LTE vendors, he says, and its Diameter signalling customers are pushing the firm to deliver Diameter solutions that scale up to 100 million subscribers or more. “’Frantic’ is a good word for how things are at the moment,” Suriano says. “The customers are pushing very hard right now, and we’re excited to be able to get this functionality into the network.”