• New analytics around network and application performance are bringing great visibility to administrators
• These new data sources rarely work with other data sources, making piecing together a picture difficult for enterprises
This is the age of analytics. Well, perhaps the computer era has always been the age of analytics, but today we are seeing a lot more discussions of analytics in general. Areas that are particularly suited to analytics are networking, as well as application performance. These areas are particularly suited because they deal with, for the most part, defined and known measurables. For instance, in networking, packet loss, latency, etc. are all known measurables. In application performance monitoring, you can measure application response time, error rates, and number of application instances among other application specific variables. There is a whole market dedicated to application performance monitoring, to say nothing of the newer observability platforms designed to help DevOps practitioners monitor performance.
Each networking vendor has its own analytics and display solutions. Most of them even include machine learning to learn and predict patterns and cut down on the number of false positives. Network administrators set thresholds for parameters such as port utilization. However, sometimes there are known and planned events that violate those thresholds. This could be an event like a weekly backup or periodic job that leads to a spike that violates the threshold, but is known and is not a problem. This leads to false positives and wastes administrator time. The learning systems learn these events and do not issue alerts, cutting down the number of alerts that have to be dealt with. Some of these systems can even attempt automatic repair of common network issues. It’s all about operational efficiency.
However, there isn’t much out there to help bring all of these data sources together. Application monitoring should be linked and crossed with network monitoring. Charts and dashboards should be able to show hybrid views, concentrating on a single application or system. Is that possible? Considering the number of APIs available in modern systems, the answer is a qualified yes. But that requires integration between two or more analytics systems, as well as data correlation. This takes a lot of time and effort to get right. To make things worse, if you have more than one network vendor, or more than one application monitoring suite – well, you can see how the work multiplies.
These products collect a sea of data, but right now stitching all this data together is more difficult than it should be. We need some standards to help with this kind of data correlation. It would be best if vendors cooperated on these standards, but if they can’t perhaps there will need to be action by standards bodies. For now, the new networking analytics capabilities are wonderful, and the world of application performance management is undergoing a revolution. But long term, vendors will need to look at analytics as a way to show their willingness to make their systems work with others, rather than as a way to lock customers into their particular solution.