In the early ‘90s the launch of MIMO delivered an enormous mobile network performance boost – effectively doubling the throughput of a frequency band. Similarly, each new cellular generation – 2G, 3G, 4G and now 5G – also contributed huge performance gains. One common denominator delivering these impressive results was Moore’s law - a dramatic increase in cost effective computing resources. Unfortunately, in each previous case, a hardware refresh was required to unlock the spectacular benefits.
Cloud computing continued the cost effective growth as predicted a few years ago in the Economist article “After Moore’s law: The Future of Computing”. The cloud has created a new opportunity to double mobile network capacity utilization - this time without the hardware refresh!
In the next several years, most data traffic growth will be in 4G as is shown in Ericsson’s Mobility Report Nov 2019. Growth happens mainly in dense urban areas where the majority of mobile users live. In those areas, this new opportunity to double capacity utilization can handle the expected 4G growth without investment in network upgrades. Thus, all available CapEx can be used for 5G.
Until now, mobile operators have basically had two mechanisms to get the best performance at the lowest cost:
- Maintain the best possible structure in their network to meet the subscribers’ needs while keeping up with continuously growing data demand
- Use the best possible control systems in the real time operation of the network
Network planning tools, fine-tuned with field data, provide support for decisions about network upgrades. Typical local area planning cycles are three to six months, and the interval between changes in any specific location is usually years. In network operation, feedback loops drive control actions within milliseconds. While self-optimizing network (SON) systems make parameter adjustments in minutes based on cell level conditions measured in real time. In addition, both planning and optimization take place within single networks. In a continuous growth environment, this requires maintaining large reserves for future growth which results in poor overall utilization.
New Domain in Network ManagementCloud computing capabilities enable optimizing across wide regions and multiple networks. This makes it possible to tap into the reserve capacity that has been built into every mobile network to cover future growth. Until now this enormous resource has been seriously underutilized. Our latest solution Xpacity employs this reserve by considering the effect of every mobile connection to the efficiency and performance across all networks.
In addition, Xpacity’s new domain of control is different from the two sets of tools in use today in terms of its scope and interval of adjustments. The solution’s scope is broader and deeper than any control or planning system. It considers hundreds of thousands of variables. The number of possible combinations, even in a single network, runs in the billions. In optimizing across multiple networks, the best combination is found among trillions of possible ways to serve the subscribers.
Finding the set of parameters that result in the best distribution of the demand across all network resources is based on massive parallel cloud computing, sophisticated algorithms, AI and machine learning. All of these help to simultaneously optimize demand and load distribution across wide regions and multiple networks.
Xpacity’s cross-optimization approach creates new degrees of freedom, which allow the pursuit of several distinct objectives that may be mutually dependent. For example, the system can strive to limit the peak loads in a region to a specified target value and, at the same time, preserve the relative competitive performance differences between networks. In addition, it can pursue both of these objectives while controlling the amount of data capacity used across network boundaries.
The other way Xpacity differs from existing tools is in regard to the frequency of control actions. The re-optimization of demand distribution is carried out every few weeks, more frequently than typical network configuration changes but without the need for a real time feedback loop.
The solution performs its calculations based on a small set of key values that are available from the network planning models, calibrated by field measurements. This way MNOs can study the network performance and load distribution improvements using their own planning tools before implementing any parameter changes in their networks.
The optimization is realized by small changes to parameters that control hand-overs between cells. While this is a narrow set of variables, the total number of possible combinations across overlapping cells, possibly in multiple networks, is very large. These calculations are what require the massive computing resources and provide the extraordinary benefits in cost and performance.
From its initial launch over 20 years ago, MIMO has evolved from doubling the capacity to today’s massive gains. Similarly, as access to more computing and cloud resources continues to grow, there will be more opportunities for revolutionary efficiency gains across the entire mobile value chain.
For now, doubling the effective capacity and performance without any network changes is available today.