Blog Post - Dave Witchalls, Aug 8 2018

How Machine Learning can help get you through Peak Trading

How Machine Learning can help get you through Peak Trading

I know that Monitoring and Alerting isn't always the most exciting of topics, but with Machine Learning becoming both more powerful and accessible, things are about to get a lot more interesting.

With a traditional setup, monitoring checks rely heavily on manually set thresholds to alert the relevant teams when things such as CPU levels, Average Response Times, order values etc.. are outside of what is considered to be 'normal'. These thresholds are based on a number of factors, from a Salmon perspective, experience of running some of the worlds largest eCommerce sites gives us a wealth of knowledge, as well as more localised experience of how well a particular set of servers cope, when for example the CPU is running at a certain percentage utilised.

This has served us well and continues to do so, however to continue to improve what we do and make sure we don't just monitor what we know about, Salmon are investing in Machine Learning to provide unattended predictive monitoring for our clients.

What does this mean?

The Machine Learning engine learns the data from systems it is plugged into, in doing this, it starts to build a picture of what is considered normal behaviour and then sets some operating tolerance thresholds. The engine continues to learn and dynamically adjust what these values should be as your site continues to grow, this includes when releases go live and any changes that might be encountered through amendment to, or additional infrastructure.

A great example of where Machine Learning is more advanced than manually set thresholds, is where you monitor for the number of completed orders. If you run a large eCommerce site with a global presence, it's not always black and white if it's an issue that you haven't had an order at 2am BST. Are your customers having an issue checking out, or is it simply that no orders are genuinely being placed at what is a fairly quiet hour. The Machine Learning engine will learn the sales behaviour of that locale and build a tolerance threshold to monitor based on an ever growing volume of data.

Another huge benefit of the Machine Learning engine is traffic trending. As the engine continues to learn about site behaviour, it can also use that data to help predict site growth based on the history that it builds. This allows Salmon to help you more accurately predict site growth and aide with capacity planning.

Machine learning, combined with our set of locally running checks, our wealth of experience in setting up some standard checks, you can rest easy knowing that Salmon Managed Services are your eyes and ears 24/7.