Editor’s Note: Welcome to our weekly Reader Forum section. In an attempt to broaden our interaction with our readers we have created this forum for those with something meaningful to say to the wireless industry. We want to keep this as open as possible, but we maintain some editorial control to keep it free of commercials or attacks. Please send along submissions for this section to our editors at: [email protected]
Wireless operators are struggling to handle explosive growth in data traffic on their networks. Tools and processes for managing spectrum, deciding where and when to add or remove capacity, technology migration planning, and approaches to merging networks largely rely on legacy methods and tools that don’t work well with newer technologies. However, recent advances in data mining (“big data”) and predictive analytics are injecting science into these processes, identifying immediate performance improvement opportunities and bringing visibility to areas of excess capacity that can be tapped easily and cheaply, maximizing the use of limited spectrum resources.
Legacy capacity planning method limitations
$100 billion worldwide is spent each year to keep pace with rising traffic, but much of this spending is misspent on new cell sites, spectrum addition or migration, and network equipment that is added too early, too late, or in the wrong place. Networks are still being planned and optimized in large part using legacy planning tools and methods, and also gut feel and intuition. Legacy tools for capacity planning often treat all base stations as having equal capacity or use simple, fixed thresholds to determine when capacity needs to be added (if traffic is forecasted to exceed “X erlangs” or “Y megabytes,” add a carrier). This approach worked with older technologies, but with newer network technologies, cell capacity and performance varies dramatically from cell to cell depending on the location of users (near? far? in buildings?), the traffic level and mix (voice vs. data), location of neighboring cells, noise from the serving and neighboring cells and users, external interference and other factors.
Existing third-party tools (such as propagation prediction and automatic cell planning tools) can predict a crude, generic quality metric over the expected coverage area, but they give no clear correlation to the performance KPIs measured by every operator and experienced directly by every customer, such as the dropped call rate, access failure rate, throughput and latency. Accuracy of typical coverage and quality predictions is degraded by the lack of precise information about the location of users, vegetation, building materials, etc. Attempts to overcome this problem by randomly distributing users through “Monte Carlo” simulations and categorizing areas into discrete morphology types improves accuracy, but still won’t accurately predict performance as experienced by customers. Operators also typically have little visibility into other real-world capacity reducers such as imbalanced carriers, cells and areas. If this data is available, it is often in table form and a great amount of effort is needed to understand whether an overloaded site is situated next to an underutilized site, for instance. New tools are required to fill this gap between theoretical predictions and the actual customer experience.
Performance-based capacity planning and optimization
Until recently there has been no accurate way to capture each cell’s unique relation between traffic and performance, or to know the level of traffic at which a particular cell’s performance will begin to degrade. However, through data mining and smart manipulation of historical performance and network configuration data, each and every carrier on the network can be characterized in terms of its relation between traffic and network performance. The operator can step six or 12 months into the future and view the predicted network performance at future traffic levels. Then network changes can be made before the customer notices any degradation due to increased traffic (a proactive rather than existing reactive approach to maintaining good network performance). Operators can now have a “heads up” on what needs to be addressed today or in the future to maintain or improve network performance, even while traffic is rising.
Once each cell’s unique “breaking point” is known, operators can identify simple, low cost ways to squeeze more capacity from existing network equipment, often highlighting large numbers of planned new cell site builds, radio adds and spectrum adds that can be cancelled or deferred. Conversely, areas that will degrade in the near future that have no relief planned can be identified and mitigated before customers experience service problems. Spectrum carving and device mix scenarios can be simulated in order to optimize timing of network adjustments and rollouts of new technologies. Networks often have excess capacity with overloaded cells located next to underutilized cells that can be proactively rebalanced through simple, low cost network changes such as antenna azimuth or down-tilt changes or power changes. By analyzing the forecasted network performance, operators can prioritize which new capacity sites to build (which can take two or more years), sites to get additional carriers (six months or more), location and quantity of small cells needed and spectrum required, and plan ahead to get these modifications done on time before KPIs are impacted. The analogy can be made to just-in-time-manufacturing: “The right material, at the right time, at the right place and in the exact amount.”
Untapped capacity and opportunities for immediate performance improvement can be brought to light in all networks through the use of new predictive analytics tools and processes. By knowing the specific breaking points in the network well in advance, network adjustments can be planned and implemented in time to preserve a good customer experience. To successfully manage rapidly rising traffic, network operators will need to adopt a performance based approach to capacity planning and optimization.