I had the pleasure of writing the following entry with Bob Sellinger, a former telecom network architect and colleague from Sun Microsystems. Bob has worked in the telecom industry for many years and served as Director of Engineering and Marketing while at Alcatel-Lucent.
In the early days of telecom network design, break even points of a million subscribers were acceptable for most business cases. Higher costs could be justified by the large subscriber base that was being enabled. Not so today. Carriers want to establish national reach while being able to cost-effectively deploy networks in evolving markets that may have only 10–25,000 initial subscribers. Carriers also want the option of deploying new differentiated services (like CDNs) with modest incremental investments that leverage their large investments in new 4G access networks. While wireless growth in major metropolitan cities is slowing, emerging markets such as the BRIC countries (Brazil, Russia, India and China) are experiencing tremendous growth in wireless adoption. The challenge for carriers is to figure out how to cost-effectively deploy wireless networks in small, remote villages where the past economics of a large subscriber base no longer apply.
The trend to deploy networks on a more granular basis, combined with the convergence of wireline and wireless networks, is prompting NEPs to seek new ways to design their network elements. Without such change, they run the risk of designing a solution that costs too much, takes too long to develop, doesn't cost effectively evolve with new differentiated services, and does not scale well on the lower end.
One path forward for the carrier is to focus on lowering the cost of their 1st subscribers by using off-the shelf technologies to radically converge parts of new 4G networks. This emphasis on low-end scalability can be accomplished with new platform technologies that support the aggressive consolidation of all the discrete network elements that typically have been deployed on physically separate servers. New multi-core processors coupled with powerful virtualization technology can perform all of the functions, including packet processing, deployed at the network’s edge with the aggregate wireline speeds (10Gb) that 4G services expect. By using a single multi-core processor capable of both running simultaneously today’s de-facto operating environments—Solaris or Carrier-Grade Linux—and the lightweight run-time environments required for processing packets at wireline speeds, the 4G order of magnitude increase in access bandwidth can be achieved in a cost effective way.
For example, next-generation IP networks are being based on the IP Multimedia Subsystem (IMS) design. IMS consists of many elements such as the HSS, CSCF (-P, -I, -S), media gateways and more. Broadband 4G networks require additional elements such as Content Delivery Networks for video transmission and network controllers such as the LTE access controller or the WiMax ASN Gateway. If the carrier uses the right platform, all of this functionality could be combined on a single, multi-core processor capable of simultaneously executing all the operating environments these functions require, resulting in dramatic cost reductions. As network traffic increases, the carrier could deploy a similarly configured blade server versus having to replicate multiple network elements.
Using industry standard 4G-platform technologies such as multi-core processors, virtualization software and lightweight run-time environments, functionality could be cost-effectively consolidated at the edge of the network in nodes that competitively scale for small numbers of subscribers. Not only would this represent a cost breakthrough for building networks for emerging markets, it would facilitate the deployment of new 4G networks on a campus or community scale. Finally, it would also create the computing foundation in the core of new 4G access networks that new services could cost-effectively exploit - to the advantage of the NEP, the carrier, and the carrier's customers.
No comments:
Post a Comment