Machine learning has crossed the chasm. In 2020, McKinsey found that out of 2,395 companies surveyed, 50% had an ongoing investment in machine learning. By 2030, machine learning is predicted to deliver around $13 trillion. Before long, a good understanding of machine learning (ML) will be a central requirement in any technical strategy.
The question is — what role is artificial intelligence (AI) going to play in engineering? How will the future of building and deploying code be impacted by the advent of ML? Here, we’ll argue why ML is becoming central to the ongoing development of software engineering.
The growing rate of change in software development
Companies are accelerating their rate of change. Software deployments were once yearly or bi-annual affairs. Now, two-thirds of companies surveyed are deploying at least once a month, with 26% of companies deploying multiple times a day. This growing rate of change demonstrates the industry is accelerating its rate of change to keep up with demand.
If we follow this trend, almost all companies will be expected to deploy changes multiple times a day if they wish to keep up with the shifting demands of the modern software market. Scaling this rate of change is hard. As we accelerate even faster, we will need to find new ways to optimize our ways of working, tackle the unknowns and drive software engineering into the future.
Enter machine learning and AIops
The software engineering community understands the operational overhead of running a complex microservices architecture. Engineers typically spend 23% of their time undergoing operational challenges. How could AIops lower this number and free up time for engineers to get back to coding?
Utilizing AIops for your alerts by detecting anomalies
A common challenge within organizations is to detect anomalies. Anomalous results are those that don’t fit in with the rest of the dataset. The challenge is simple: how do you define anomalies? Some datasets come with extensive and varied data, while others are very uniform. It becomes a complex statistical problem to categorize and detect a sudden change in this data.
Detecting anomalies through machine learning
Anomaly detection is a machine learning technique that uses an AI-based algorithm’s pattern recognition powers to find outliers in your data. This is incredibly powerful for operational challenges where, typically, human operators would need to filter out the noise to find the actionable insights buried in the data.
These insights are compelling because your AI approach to alerting can raise issues you’ve never seen before. With traditional alerting, you’ll typically have to pre-empt incidents that you believe will happen and create rules for your alerts. These can be called your known knowns or your known unknowns. The incidents you’re either aware of or blind spots in your monitoring that you’re covering just in case. But what about your unknown unknowns?
This is where your machine learning algorithms come in. Your AIops-driven alerts can act as a safety net around your traditional alerting so that if sudden anomalies happen in your logs, metrics or traces, you can operate with confidence that you’ll be informed. This means less time defining incredibly granular alerts and more time spent building and deploying the features that will set your company apart in the market.
AIops can be your safety net
Rather than defining a myriad of traditional alerts around every possible outcome and spending considerable time building, maintaining, amending and tuning these alerts, you can define some of your core alerts and use your AIops approach to capture the rest.
As we grow into modern software engineering, engineers’ time has become a scarce resource. AIops has the potential to lower the growing operational overhead of software and free up the time for software engineers to innovate, develop and grow into the new era of coding.
Ariel Assaraf is CEO of Coralogix.
The post Why AIops may be necessary for the future of engineering appeared first on Venture Beat.