Network performance and visibility: The art of KPIs

The network today has transformed considerably from what it was a decade ago, and its transition will continue. It has evolved from being well-defined to perimeter-less, from largely manually supervised to mostly autonomous, and from IT-oriented to business-delivery-oriented.

During the course of these changes, it is critical for network managers to have a firm grasp on the state of the network, so they can ensure that it is functioning optimally and delivering the business services for which it has been deployed.

But in the era of digital transformation, the enterprise WAN is unable to keep up with demand. There are significant challenges with on-going migration to the cloud and business agility.

These concerns underscore the critical importance in securing the right network infrastructure and setting the right performance metrics to ensure network transformation efforts meet customer expectations. The right KPIs or Key Performance Indicators will lead to network connectivity improvements and application performance solutions.

Once a KPI baseline is set, any metrics crossing a certain threshold can automatically trigger alerts or tickets; while these alerts can be handled by either a human network manager or an automated self-correcting system, the important aspect is putting in place a regime that defined the boundaries of acceptable performance. To this end, setting realistic and measurable KPIs is necessary.

The KPIs will define parameters for some of the most important quantifiable metrics a network is able to output: availability, jitter, latency, utilisation and workload, among others. Not only is the metric important at a specific given point, but in the age of elastic workloads and devices registering on and off the network at short notice, KPIs also need to take into account historical trends of the network’s behaviour. Current and historical data can be extrapolated to predict network behaviour and even adjust KPIs if they are deemed too stringent or liberal.

Service Level Agreements

Service Level Agreements (SLAs) is the most direct and obvious reference to set KPIs with, whether agreed upon with an external service provider or internally, with business users. Certain tools are able to directly measure KPIs against SLAs, making the process of meeting standards self-regulated. The vital aspects of network performance, such as availability (the percentage of time the network is functioning) and latency (the delay in data being transmitted across the network) are often defined within the SLAs, making setting KPIs convenient, and directly accountable to business value.

Where and when to measure?

The exact points at which performance is measured is also meaningful in getting accurate insight into the state of the network. For example, setting measurement points at the entry and exit can provide a picture of the latency the network is subjecting data to, as well as packet loss the data is experiencing. Both measures are crucial and can have great bearing on meeting KPIs.

Similarly, the frequency of measurement can be decisive, for metrics such as jitter (variation in time between data packets, often caused by congestion). Because of the fluidity of the modern network, jitter should be measured at regular intervals, and the KPIs should reflect this. Likewise with latency, which can change constantly across too-long a time frame.

The need for feedback

Dashboarding tools help businesses gain clear understanding of network KPIs, which can have an impact on business decisions. For instance, for a network suffering from availability issues, mission-critical business workloads can be avoided, but for purposes of regular backup or archiving functions, latency issues will not have a major impact.

A toolkit or a dedicated system for managing KPIs can summarise, visualise and sometimes even predict network performance for decision-makers, simplifying the work of the managers. To achieve this, the dashboard should provide true visibility into the application layer to match network resources to performance KPIs.

This visibility serves as a data-driven window into multi-cloud deployments, application delivery, network behaviour and dependencies that will help map out optimal network paths.

With feedback and visibility, KPIs become relevant not only to the network managers and their immediate stakeholders, but also for long-term evolution of the network industry. KPI reports can be compiled and analysed to trace the overarching changes in iterations of networks, and can be an invaluable metric for network designers to take into account when working on subsequent software or hardware updates.

KPIs in an automated world

The network performance management and KPIs will remain imperative in a world where intelligent, self-healing networks might be a reality soon. The decision-maker may change, from human to robotic; but ultimately, clear and transparent measures of the network’s state of health will be the main data points against which decisions will be made by whatever entity is responsible for managing the network. Robust KPI regimes and standards set in place will ease the transition towards the networks of the future — the greater the automation, the greater is the need for visibility into network performance and for hard, quantifiable metrics to maintain control over a network.

Find out more at www.singtel.com/liquidinfra or speak to us.