Network users need consistency.  A network where response times vary from lightning-fast to “go get a coffee while the screen loads” is one guaranteed to bring complaints.  Networks are at the core of most business operations. Slow, unresponsive networks can reduce business efficiencies and prevent them from attaining an edge over their competition.

Networks are becoming more and more complex and maintaining an equable level of service to users more and more vital.  Network underperformance or failures can be very damaging to a business, and sometimes catastrophic. The continuous availability of online retail operations, for example, in online stores and airline reservation systems is vital to sustaining a business.

The causes of network instability and failure will vary and are often difficult to pin down. It could be a network cabling or equipment fault.  An application may have complex screens with lots of graphics and background processing. Someone might be processing a large print job to a network printer.  It might be an incorrect desktop or network configuration or the result of a recent change in network architecture.  It could even be the fault of a third-party service supplier.

Network management needs to investigate network issues and how to rectify the problem.  That is where Network Analytics comes in.

Network Analytics

At it’s most basic, network analytics is the analysis of network performance data to determine trends and patterns. As a bonus, the measurement of normal patterns of usage eases the identification of external threats.  DDoS attacks can be detected at an early stage by sudden changes in traffic type and pattern. Network management can then take any actions necessary to prevent or fix network problems.   

In general terms, network management prepares models of how they expect the network to perform, and compare actual performance against the model.

It can be a manual process; for example, using spreadsheets; software applications can assist, or there is an analytics function in the network architecture software in Software Defined Networks (“SDN”).

The current trend is to move to an SDN, but in the meantime use intelligent automation to try to lessen the load on network admin.   AI and machine learning are proving to be critical components of the new wave of network analytic tools.

It is generally a four-step process, following the recognition that there is an issue:

• Problem Definition; The problem is identified, measured, and set out using performance data taken from network monitoring, and the perceived effect on users.

• Remove variables; Eliminate any temporary issues that could be causing the problem.  For example, a network route might be unavailable during switch replacement.  

• Find the root cause; and as Sherlock would say, having eliminated all the variables, whatever is left, however improbable is the root cause. 

• Fix it. Easy to say, perhaps not so easy to do.

Current trends

Network Analytics is a relatively mature science but has been dramatically improved recently by the addition of Machine Learning and Artificial Intelligence to the mix.   The root cause of these improvements has been the development of SDN networks on the way to Intent-Based Networks.  In both network models,  the bulk of the detailed configuration and management tasks are carried out by software.

Machine Learning and AI offer significant increases in the speed of providing accurate insights and preparation of solutions by the analytics engine in a dynamic environment.  Performance models are modified on the fly as network configurations change.

Another driver is that applications processing high volumes of data with a short time constraint, such as driverless vehicles and IoT management in a manufacturing environment need instant rectification of network problems.

Benefits of Network Analytics

There are both cost and productivity benefits to network Analytics.

• Cost benefits

  

  • New Revenue Streams

    In some data-driven environments, analytics has a significant place in generating insights into the potential for adding new revenue streams to an existing environment  

  • Quicker to Market

    Again in data-driven or hosted service suppliers,  analytics makes it much quicker to carry out capacity evaluations when considering adding new clients or new or expanded services.  Time to market is significantly improved.

  • Optimising Resource Usage

    Analytics provide a deeper understanding of how the network operates and reacts,  helping management to optimise the deployment of resources in a balanced manner, reducing operational and development costs.

• Intangibles

  

  • Productivity

    The network management staff don’t need to wade through reams of network usage data to identify a problem and the best fix.  Analytics will do that for them and continuously, leaving them for other more beneficial tasks.

  • Malware Protection

    Significant changes in the volumes and types of network traffic often signal a malware or hack attack.  Analytics can identify these changes almost immediately, providing an early warning and additional time to remedy whatever is happening.

  • Continuous Improvement

    The automatic analysis of performance data gives network management valuable information for capacity planning and assessment of changes and extensions to network architectures.   This will foster a culture of continuous improvement.

While Network Analytics is often seen as the Cinderella of network management, it is a vital component in keeping a network operational.