They’re accustomed to making decisions on the basis of installing infrastructure for rolling out a single service, according to Marc LeClerc, service layer business advocate for Ericsson AB (Nasdaq: ERICY), one of the speakers. That doesn’t make sense when operators are installing SDPs, which enable them to roll out almost limitless numbers of services, often provided by third parties, over converged IP networks.
“We’re trying to provide a killer environment rather then a killer application,” LeClerc noted in the conference’s first session, “Service Management: Linking SDP Intelligence with Business Success.” “We have a challenge as an industry over how we change the way we analyze the return on investment of SDPs.”
The single-service-per-platform mindset is preventing some operators from realizing that SDPs have low upfront costs compared to traditional systems and that the economics get even better as more services are rolled out, the conference speakers said.
“The skills needed to develop these kinds of services are open-type skills,” said Derek Howlett, director of mobility practice at Sun Microsystems Inc. (Nasdaq: SUNW). As a result, “SDPs are going to be a lot cheaper to build" than the infrastructure used to support traditional services.
The same point was made to Light Reading by another speaker, Dan Lowe, chief strategy officer of Tazz Networks Inc., prior to the start of the conference. Lowe was asked to comment on today’s news of AT&T Corp. (NYSE: T) picking Microsoft Corp.’s (Nasdaq: MSFT) SDP (see AT&T Adopts Microsoft's SDP). He replied that 10 million programmers around the world were familiar with Microsoft programming languages, and that gave the software company a huge advantage in terms of enabling operators to drive down costs.
The ability of SDPs to support hundreds of services reduces costs even further, noted Ralph Corey, director of service provider solutions for Intel Corp. (Nasdaq: INTC), another speaker on the panel. “Traditional systems don’t scale that well,” he said, "and really, after four or five services, they don’t scale at all."
Operators could roll out hundreds of services using an SDP, thus spreading costs much more thinly, Corey said. However, even SDPs couldn’t scale indefinitely. Operators would likely need to add infrastructure and separate functions when the numbers of services reached several hundred. A couple of slides in the conference session summarized an exercise conducted by Intel, to quantify the costs of deploying an SDP for 30 million subscribers, 2,250 of them being active in any one second. The hypothetical SDP supports four services:
- Location-dependent personalized content delivery with multi-device rendering for 60 percent of the transaction. A total of 1,567,998 transactions per hour are supported, with a response time of less than 700 milliseconds.
- Media downloads such as video and audio files and Java-based games. A total of 389,880 transactions per hour are supported, with a response time of less then 900 milliseconds.
- User change profile, enabling subscribers to have calls find and follow them. A total of 285,768 transactions per hour are supported, with a response time of less then 1,100 milliseconds.
- SIP Call Control: A total of 7,560,000 busy hour call attempts and 2,100 call setups per second are supported, with a call setup time of 243 milliseconds (95th percentile).
The overall cost of the servers and software, based on list prices, is $10.8 million. This equals $0.0136 per user transaction and $0.3605 per subscriber. Corey says that costs for a traditional system would be measured in dollars rather than cents. He also says the costs in real life would be lower, because carriers could negotiate discounts off the list prices used by Intel. Rolling out additional services using the same platform would reduce costs as well, although there would be diminishing returns when large numbers of services were already deployed, Corey added.
Intel’s exercise has been used by a number of carriers to evaluate the business case for SDPs, according to Corey. Those carriers have plugged their own numbers into Intel’s study and have gone ahead with implementation projects, he told Light Reading.
Ericsson’s LeClerc pointed out that systems integration costs typically double the overall bill for operators, and that systems integration can account for an increasing proportion of the overall bill as the number of services and subscribers increases.
At the other end of the scale, Intel’s Corey says that service-providers would only have to pay around $50,000 to $100,000 for an entry-level SDP supporting about half a million subscribers.
That still seems to be more than many service providers appear willing to pay. The conference moderator, Heavy Reading senior analyst Rick Thompson, gave a few highlights of a survey of 65 service providers in his introductory remarks. It indicated that 22 percent of operators aren’t prepared to pay more then $10,000 for an SDP, and a further 24 percent don’t want to pay more than $100,000.
The survey also indicated that “high cost of deployment" and "uncertain return on investment" rank as the biggest barriers to operators taking the plunge with SDPs.
— Peter Heywood, Founding Editor, Light Reading