Dr. Hossein Eslambolchi
January 2012
I have some experience in packet-based network design, including quite a bit of time working on wire-line and wireless networks over the past decade. Given massive changes occurring in the industry — especially in wireless technology — I humbly offer some high-level design principles that I believe need to be incorporated into the design of any packet-based network. The main rationale for these principles is the massive growth of data over the next decade.
In PSTN was defined [measured?] by a very helpful model called erlangs of traffic. All switches and networks were designed based on the number of erlangs of traffic using poison distribution. But given the unpredictability and non-deterministic nature of data networks, one has to re-think how the packet networks of the future should be designed.
Broadband networks are primarily designed for web surfing and consumer streaming, like video. Given this general use, consider these rules:
1. For each percentage increase in edge capacity, core capacity needs to grow by 30 times. This is unsustainable with the current pace of innovation in the industry; a new edge and core architecture needs to replace traditional architectures from the early 2000s.
2. Host content like 3D, HD and 4D video has arrived or is on the horizon. The new video distribution format would be best for ISPs and one can use same model for mobile content as well. Although Mobile content is more complex, where CDNs ecosystem need to be very sophisticated and ISPs need to use packet layer data to determine where contents be hosted closer to customer location.
The demand of traffic growth alone is creating many challenges for mobile network operators. For example, spectrum efficiency could be boosted by a factor of six by 2018. However, before more spectral efficient systems become available, one needs to off-load at least 2 times the current 2011 level of packet traffic. I believe that by 2018-2020, 7 percent of traffic will be national mobile data and Current Cellular networks will end up driving at least 15 percent of cellular traffic off-load. Clearly, devices like the Amazon Kindle, smart phones, and new tablets from other vendors will make the problem even worse. The challenge is to manage large-scale data networks with much better unit costs and efficiency.
As an IP technologist, it’s my belief that broadband off-load can potentially eliminate 2-5 GB of traffic intended for media-hungry devices each month. Additionally, broadcast technologies of the future are not limited to live TV — smart devices are able to cache and retrieve programming.
Through off-loading, we can reduce congested cellular traffic by at least 50 percent. Only a small amount of bandwidth is truly needed in that scenario. I assume about 5 percent of savings will be sufficient to subsidize extra device costs. ISPs may also have to provide highly affordable delivery for content owners and create smart monetization across their networks.
So, how do these design choices benefit mobile network operators or ISPs with worldwide mobile networks?
1. They reduce capital and operating expenditures by delivering content locally;
2. They enhance network performance for content delivery; and
3. Leverage content across multiple screens such as smart phones, tablets and appliances with Internet connections (including automobiles).
I see a great opportunity for MNOs and CDNs to support of customer needs by leveraging:
A. Subscriber information
B. Roaming state, geo-location information analytics
C. Network state, congestion and user sanity checks
Under this scenario, traffic steering can be managed much more intelligently than it is today. We can control the traffic conditions, along with application-level routing decisions, based on MAC, user agents, language, cookies, data and other types of info MNOs possess.
Opportunities for mobile operators to exploit new streams of revenue still exist. In fact, there are huge revenue streams available. Simple principles like the ones I sketched above are still innovative — even revolutionary. I call this approach smart revenue monetization; I believe that it must be adopted by service providers if they are to succeed in the 21st century.
Dr. Hossein Eslambolchi
January 2012
