Selecting the best Network Intelligence strategy will help operators meet the needs of rapidly changing business models, writes Keith Cobler
Network operators today are faced with many challenges that are impacting their businesses, some of which include increased market competition, technology challenges and a more demanding customer base, to name a few. Many operators have been blind-sided by these and other industry challenges and are searching for viable solutions. Business models of the past no longer work in this new market environment. Operators today recognize the importance of running their businesses based on actual data that can be analysed and evaluated in real-time. To do this, network operators are relying on network intelligence solutions that correlate high-level business objectives with what is actually occurring at the network level.
Different from Business Intelligence solutions, Network Intelligence solutions start by capturing the raw data and events that take place at the network level and transform this data into actionable information. They rely on the collection of real-time, high-quality events that transverse multiple network technologies as the basis from which meaningful information can be obtained. To some degree, the old adage of ‘garbage in – garbage out' applies since the quality of information at the output is very much dependent on the data-in and how that data is transformed into information.
In general, Network Intelligence solutions consist of three integrated components: (1) collection agents that collect data from network interfaces and elements, (2) a network correlation layer that ties together related network events and (3) analysis capabilities that empower multiple departments within a network operator with key information that allows them to manage better their individual departments and their business as a whole.
In terms of solutions, there are many solutions on the market but few that cover all three components of collection, correlation and analysis. Based on this, operators who implement a Network Intelligence strategy may decide to piece together the top-to-bottom solution using different vendors, or will look to a single provider to supply all the necessary components.
From Data to Information
A Network Intelligence solution needs to collect events at the network level and then correlate or tie them back to specific business objectives or desired outcomes. Within the network, events and transactions are captured and stored as a call data record, or CDR, which is generally used as the primary data source for calculating key performance indicators (KPIs) by network, service or customer assurance or other OSS/BSS application.
Today's Network Intelligence systems rely on signalling and media data as the primary data sources from which intelligence can be derived. Some Network Intelligence systems may also use data supplied from network elements or from other OSS/BSS systems. Different network technologies and topologies also play a factor in the types of data that can be collected and used. As networks continue to evolve, the trend has been towards combining elements and functionality into as few elements as possible in an effort to reduce costs. From a Network Intelligence perspective, this has created some new challenges in terms of getting access to the data; e.g., in LTE networks, a new network element called the E-NodeB (Enhanced Node-B), is a combination of the Node-B base-station and the RNC as a single element, thus physically eliminating the Iub interface.
Data collection devices for Network Intelligence solutions consist of probes (passive or active), element feeds and software agents. The primary difference between passive and active probes is that passive probes are non-intrusive, meaning they do not interfere or insert themselves into the data path, but rather capture the data using a mirrored port. Active probes, on the other hand, inject a test signal into the network and then measure the response of the network to that input. Software-based collection agents are generally used when physical probe deployment is impractical due to size or cost constraints.
The second key layer of a network intelligence system is the correlation/mediation layer. The purpose of this layer is to correlate all data sources end-to-end across the network and then to write this data to a CDR for post-processing. Doing this is not as simple as it sounds since most networks today are not based on a single, homogenous technology but rather have evolved and consist of a patchwork of legacy and next-generation technologies. And to correlate a call or session end-to-end across multiple network technologies requires a sophisticated protocol correlation engine that can piece together the protocols across every leg of the connection, or in the case of IP, derive this correlation from the IP packets themselves.
After the data has been collected and correlated, the last layer of a network intelligence system is the processing of the data into meaningful, accurate information that can be used by different individuals and organizations within the carrier.
At the heart of information analysis is the KPI, or key performance indicator. In order to correlate accurately a desired business outcome with events that occur within the network, considerable attention must be paid to identify and define KPIs correctly that are meaningful and accurate and indeed drive desired business results. This is never an easy task given the underlying complexities at the network level and the interdependence between events and variables that describe these events. For Network Intelligence solutions, it is the attention paid to modelling a desired outcome accurately and efficiently, which requires an in-depth understanding of what and when to measure, that defines the value of that system.
Different from counter values that describe a given state, KPIs are formulae that give greater insight and are based on multiple inputs such as cumulative counter values, constant values, timer values and even other KPIs that have already been computed. It would be a mistake to compare one KPI to another just by name alone without knowing the exact definition of the KPI, the criteria used to select its inputs, and any other information that may impact the KPI's accuracy or manor in which it can be used.
Linking your desired business outcome to events at the network level is the basis of network intelligence. To do this, a model of the system needs to be developed that links the desired business outputs to the dependent variables at the network level. In most cases, model development is a science in itself with many considerations that are beyond the scope of this article, but in general, the process consists of three basic steps: (1) model the system, (2) compare – modelled vs. actual and (3) optimize the model. Some network intelligence systems that are available today provide off-the-shelf KPI packages, which others provide the capabilities for users to create and modify their own. In most network intelligence implementations, it is usually a combination of the two approaches that provides the most cost-efficiency and flexibility.
Another key consideration at the analysis level is the ability to provide information in real-time to those individuals or departments who need it. In addition, the format and display of this information should be tailored to the individual groups or individuals that use it. For example, whereas network operation teams may receive their information in the form of real-time dashboards and alarms in the NOC, product planning teams may require a dashboard or report that looks at more historical or geographical trends. The point being that although a Network Intelligence system leverages a common data stream, different departments within the network operator require to see the information in a format that is most beneficial and meaningful to them.
What to consider
In general, there are two classes of network intelligence solutions available today. The first class of solution is the vertically integrated solution that contains all three functions of collection, correlation and analysis and is based on an open architecture. These vertically integrated solutions are usually provided by a single vendor, but at the same time are designed using open architectures with the required hooks required to support 3rd party hardware and software. The second general class of solutions is referred to as partial or point solutions because they tend to focus only on a single layer of a Network Intelligence solution and not the entire top-to-bottom integration you find in vertically integrated solutions. Although it is a bit of an apples-to-oranges comparison, vendors of point solutions will often sell their customers that their hardware or software components can be integrated with other 3rd party hardware and software.
The chart above includes a summary of the two general classes of Network Intelligence solutions that are available today. For each of the three components that make up a Network Intelligence solution, listed are some of the key attributes that define each of these. Depending on your specific circumstances, this list may be expanded or modified considerably – and is only intended to be a rough guide for comparison.
Ultimately, the important points to keep in mind when selecting any type of Network Intelligence solution are:
1 Does the solution accurately and effectively correlate my desired business outcomes to what is occurring at the network level?
2 Does the solution provide an actionable path for different groups or individuals within the company to identify an issue and take action that has a positive business impact?
3 Does the solution have a justifiable Return-on-Investment (ROI)?
Summary
Network operators today are faced with tremendous challenges. In order to reduce the risk and better ensure that their business objectives and strategies can be achieved, network operators are moving towards Network Intelligence solutions as a means to achieve their business objectives based on what is actually happening at the network level. In short, network intelligence solutions consist of three components: (1) collection agents – that collect data from the network, (2) correlation layer or engine – that ties together relevant data from across the network and saves it in an accessible data file and (3) analysis packages – that turn the data into meaningful information that can be used by multiple departments within the network operator. The two general classes of solutions available to the market today are vertically integrated solutions and point, or partial solutions. Determining which type of solution best meets your needs is not simple, and comes down to decisions related to budget, resources and what types of systems and technologies you currently have in place. But the bottom line is that for whatever Network Intelligence system you choose to go with, it must have the ability to drive positive business results based on accurate and real-time information.