Put away your tarot cards – big name carriers are discovering the power of predictive network testing. Automated Tests that monitor signal attenuation in fibre and warn of failure before it happens promise savings in time, cost and reputation across the telecoms industry says Daryl Cornelius Director of ITO at Spirent Communications.
Network monitoring is becoming highly sophisticated. When something goes down, today’s systems can provide precise data on the nature, time and location of the fault – informing support staff where to go and what repairs are likely to be needed. In a stroke this reduces time to diagnose faults reducing the downtime. But it is still a failure, however fast the responses.
Latest developments in network automation however, deliver foresight – not hindsight! These developments can keep track of signal strength (Optical Degradation) – they note increasing signal attenuation points and map the rates of change to enable the prediction of laser failure! The automated testing on the automatically discovered network links, based on trigger points from decay statistics, warns in advance of likely failure time BEFORE the failure! The same maintenance job now takes place – but when you want it to, in your schedule, and without the disruption of a link failure.
It sounds simple in those terms, but bear in mind the complexity of today’s networks. This is not simply a question of measuring signals at each end of a single fibre link, but rather of multiple measurements taken at a number of points in the network by a test system that has built up a model of the entire network architecture from which it can trace signal changes and react to work in those areas of most interest, enter automatic network topology discovery!
A key factor in this process is the automation of the test creation and operation: the manual labour needed to attempt such a complex process across millions of connections without automation would make it well-nigh impossible. In fact, not even the network operators can be expected to have a full picture of today’s complex and fast evolving networks, as the key to this preventative maintenance is the ability to automatically discover the network topology. Also bear in mind; virtualization needs a very complex test to be created piloted and fine-tuned off-line, to allow subsequent faster roll-out and implementation.
This system is now being rolled out in real life. One of the world’s largest tele-communications companies operating in 25 countries worldwide has been using Spirent’s iTest integrated test creation and execution system for while. They launched the first pilot predictive testing solution based in their Network Operations Centre (NOC) and are already seeing the benefits across their optical network, measured in $ return.
No longer just fire-fighting failures as they happen, the company gets advance warning of laser failures and can plan support staff schedules in advance for greater efficiency as well as avoiding down time. Of course the network does have high redundancy to avoid network downtime, but service performance can still be compromised during failover, so predictive monitoring allows leeway for choosing the optimal time for network change.
Another leading mobile phone operator is claiming 15% better customer experience – in terms of a range of quality indicators including: availability, accessibility, retain ability, integrity, and mobility – as well as 95% accuracy in identifying potential failures from their predictive operations.
The point is that, given automation tools working in harmony with incumbent Network Management the NOC can make sense of real-time data from the network; this has of course, important implications for any highly critical network. Identifying changes in signal strength is one example of turning information into Knowledge & Action. This example of a pattern that could either sound a warning to the operator, or trigger an automatic failover to a redundant path or help remove repetitive manual tasks.
Add to that the expertise of network test professionals to suggest and develop suitable data mining algorithms, and you gain extraordinary powers to predict future performance.