November 2017 Summaries

Harness effectively utilizes its Continuous Delivery (CD) platform to manage software deployments, achieving a high success rate of 94% across numerous deployments, while employing machine learning for anomaly detection to preemptively identify and resolve potential issues before they reach production. The company practices what it preaches by using its own CD-as-a-Service platform, incorporating extensive customer feedback to drive continuous improvement and innovation. Recent deployment data reveals an increased number of failures, which are considered beneficial as they prevent problematic releases from impacting customers. A detailed examination of a specific failure highlights the importance of rigorous testing and verification processes, which include multiple stages and the use of advanced algorithms to detect anomalies in unstructured data. This approach underscores the critical role of quality assurance in ensuring the reliability of new builds and the proactive identification of issues such as the NullPointerExceptions related to a newly introduced Encryption service.

Integrating AI and machine learning (AI/ML) into Continuous Delivery (CD) pipelines streamlines the process of verifying production deployments by automating tasks traditionally done manually, thus enhancing reliability and performance monitoring. Harness employs unsupervised machine learning to automate the verification steps in deployment pipelines, drastically reducing manual effort and allowing organizations to better understand the business impact of their deployments. By leveraging data from tools like AppDynamics, Splunk, and New Relic, Harness enables real-time analysis of key performance indicators (KPIs) such as business revenue, application performance, and resource utilization. This AI-driven approach identifies performance and quality regressions, such as anomalies or failures, and can initiate automatic rollbacks, ensuring deployments are both efficient and reliable. Additionally, developers can provide human feedback to improve the accuracy of the machine learning models, making the process more precise over time.