SON Load Balancing and the Success of Disruptive Ideas

By | August 19, 2013

disruptive ideasThirteen years ago we were working at Metawave to address the capacity challenge operators were facing (sounds familiar?). Our solution was an adaptive system that automatically adjust the antenna beamwidth to change the coverage area of a given sector according to the traffic load. By changing the antenna patterns, and consequently the coverage footprint, it is possible to shift traffic from one sector to another. In other words, we were doing ‘load balancing.’ This ought to sound familiar too, right?

Load balancing is today one of the fundamental tenets of Self-Organizing Networks (SON). SON came about through the efforts of multiple organizations. The Next Generation Mobile Networks alliance (NGMN) published its requirements for SON in 2008.   This was followed shortly by 3GPP activities introducing SON in Release 8 LTE specifications in December 2008 focusing on auto-configuration. Subsequent 3GPP releases enhanced the widened the scope of SON.

Today, antenna features like RET (remote electrical tilt), RAS (remote azimuth steering) and RAB (remote azimuth beamwidth) allow operators to control the coverage footprint of a site. These features can be combined with SON load balancing algorithms to dynamically optimize network performance. This is not as simple an exercise as it sounds because there are many interdependencies, but we will get there. When we do, we’ll have an elegant and cost effective solution that solves some important capacity challenges while at the same time it will reduce cost because automation of the network driven by different SON functions.

I say elegant and cost effective, because this was the main problem with our solutions 13 years ago. Back then, the system was an add-on to existing base stations (in other words you had high-power radios for the base stations and another set for the add-on antenna system doubling power consumption and cooling requirements). It required a whole rack per sector (think of the consequences to space in cell sites that were crowded with big, bulky 2G systems and smaller, little less bulky 3G ones that were starting to be deployed at that time). Many RF feeder cables connected the active intelligent system with the antenna. In comparison, today’s base stations will implement such functions natively, based on standardized processes and interfaces.

SON load balances capacity in congested LTE base stations (source: NSN).

SON load balances capacity in congested LTE base stations (source: NSN).

Reflecting on this experience I can’t help but point out a couple of points. The obvious one is that capacity was always the problem with wireless systems and will always remain to be the issue. It’s a raison d’être of many companies and technologies no matter how absurd the idea might be (some might call them ‘pioneering’ ideas; and don’t get me wrong, I am not against such ideas: in fact they are necessary for evolution. Just remember that many, if not most, will die off).

But the important point is that when it comes to infrastructure, new technologies will see good traction quickly when the incremental cost to implement them is low. Take an example: beamforming vs. MIMO. Beamforming as a concept has been around for longer than MIMO but it’s different forms saw little deployment traction whereas MIMO came around in the mid 1990’s and was very quickly adopted first by 802.11/Wi-Fi, followed by 802.16/WiMAX and then in LTE. Beamforming is still trying to carve a niche. The elegance and relative simplicity of MIMO coupled with low incremental cost made it an attractive solution against beamforming solutions. Base stations were always deployed with two antennas. The only difference was that enabling MIMO required only one additional transmitter (and of course the logic to drive MIMO algorithms). The concept of ‘incremental cost’ is not only financial, but applies equally to process, behavior, and other factors that regulate our activities.

Going back to our load-balancing beamformer, it was an idea well ahead of its time that saw limited deployment but not enough for the company to make a business out of it. The company did not survive the dot-com recession and its assets, mainly patents, sold off at a large discount. The idea was too great of a disruption from different perspective including cost and practicality.

There are many ideas today to common problems that have been around for a long time.  Many of these ideas are not new, but how they are implemented has evolved. Some think they will succeed this time around because ‘today it’s different.’ But when it comes to wireless access networks, the outcome will not be different if the idea represents a major disruption.