Monitoring a business means monitoring an entire business – not just IT or application performance. If businesses truly care about differentiating themselves from the competition, they must approach monitoring holistically. Separate, siloed monitoring systems are quickly becoming a thing of the past.
I see time and again cloud monitoring companies working with a myopic focus on the Infrastructure area – a critical mistake. They concentrate on system health but avoid business health like the plague. Although CPU, Disk, Memory and other infrastructure KPIs are essential to maintain a healthy system, their coverage is limited and lacking an equally crucial component that drives how well a company is operating – its business. Today there is simply no excuse for having incomplete monitoring capabilities, and it is more necessary than ever to get out of monitoring siloes.
Cloud Monitoring 1.0 and the Evolution of Metrics
Monitoring infrastructure provides some visibility to overall system health by keeping machines up and running – but it is not at all adequate to determine what is occurring on the business side of a company. Infrastructure monitoring is also far too basic to keep up with updates within applications – essentially putting blinders on a company's leadership.
As it stands, infrastructure monitoring tools usually run in conjunction with other internal tools to gain an angle on the business, or analysts rely on Business Intelligence solutions that may be connected to infrastructure monitoring through internal scripts. In most cases, these 1.0 level tools require a great deal of internal development and maintenance which are difficult to scale.
In the past few years, time series metrics have been the main driver of growth in cloud monitoring systems. This approach of normalizing almost all data per a single time series representation has enabled the provision of generic solutions for many cases and different customers. Because of its rudimentary ability, it is not surprising that open source solutions are becoming so widespread among the businesses which are beginning to understand the importance of monitoring. The ability to represent all metrics in the same manner using the same dashboards and time series function sets has significantly simplified this monitoring method providing good but not fully comprehensive information.
Today's Challenges of Monitoring Business
One of the main challenges of monitoring business KPIs is that static rules and alerts are too limiting. Particularly for metrics that change per trends or seasons, static alerts are difficult to maintain because of their inherent variability. Even in the simplest cases, it is very difficult to define thresholds for thousands of metrics because it requires the user to have working knowledge of their normal range. For e-commerce companies, the holiday season is always a peak time in sales and every metric is going to behave "abnormally." It is nearly impossible for large data-driven companies, which are monitoring so much, to start making changes to reset the threshold for every single metric – talk about a nightmare.
Another challenge of monitoring so many metrics is defining rules manually especially when each metric has a different normal range. Unfortunately, it is essential that this be done to achieve effective configuration. Amazon needs to know that "Elf on a Shelf" dolls are going to sell heavily in November and that gift certificates will be sold later in the month.
Cloud Monitoring 2.0: for IT, applications AND BUSINESS
The newest generation of monitoring centralizes all company activity into a single unified solution, rather than separate solutions for IT, application, and business. This is the holistic understanding that companies have been working towards for so long – the ability to understand every metric separately and together. It is one thing to see an infrastructure anomaly on its own, but to be able to contextualize it with the correlated impact on the business affords an entirely new way to problem-solve and measure the health of a company. Beyond addressing the immediate issues this type of top-down monitoring approach offers tremendous value.
Without a smart mechanism to monitor so many rules and alerts, companies are bound to compromise what they monitor, sacrificing all for a few selected metrics. Analysts are not fortune tellers – there is no way to define what the best metrics are to monitor. This creates an inevitable delay in detection of issues, which severely limits how proactive a company can be in the varied business scenarios it faces. It also limits the granularity of the organization's visibility – bringing us back to where we were with Cloud Monitoring 1.0.
Only recently the implementation of AI in BI is enabling companies to solve challenges in monitoring. By automating the ability to differentiate between what is normal and abnormal behavior (no matter the trend or time of year) businesses finally have a chance to review a comprehensive and automatic evaluation of anomalies. With the addition of AI to monitoring, companies can differentiate themselves by how quickly they respond to changing conditions; how quickly they find bugs and glitches, how rapidly they respond to customers in crisis, and how swiftly they leverage a business opportunity triggered by a celebrity's viral Instagram post.
While companies engage with their customers in more ways than ever before, finding ways to break out of monitoring silos is going to be the key that companies use to successfully scale and compete with industry giants.
David Drai is CEO and Co-Founder of Anodot.
