August 2019 Summaries

Harness has introduced several new features and updates to enhance its Continuous Delivery platform, including tag management for better pipeline organization, support for Cyberark Application Access Manager to bolster security, and a new Essentials pricing tier aimed at smaller developer teams. The platform's authentication setup has been streamlined, allowing for a more simplified user experience by merging SSO Provider and Login Settings into a single screen. Additionally, Harness now enables deployment to multiple AWS accounts using the same delegate and has improved its audit trail with a new filter feature for easier tracking of account changes. The company is also engaging with the community through various events, meetups, and webinars, offering customers the chance to share their unique tips and tricks for a reward.

Microservices architecture, gaining attention alongside cloud migration, serverless, and Kubernetes, offers flexibility and independence in software deployment by breaking down applications into smaller, modular services. While this approach reduces dependency complexity, it increases deployment complexity, necessitating robust pipelines to manage the surge in services, as exemplified by tools like Harness. The architecture promotes more granular and independent deployments, allowing teams to work without bottlenecks, but it also introduces challenges such as increased service management complexity and potential endpoint proliferation. Despite these complexities, microservices represent an evolutionary trend in software design, providing a contrast to the monolithic architecture where changes in one functionality require entire application redeployments. The adoption of microservices involves a shift to systemic complexity, where the number of deployments increases, and infrastructure responsibility shifts towards application teams. This architecture necessitates strong deployment pipelines to avoid bottlenecks and ensure the efficient release of new services.

AWS Fargate simplifies container management by removing the complexity of Kubernetes cluster infrastructure, offering an alternative for deploying containers without the need to manage the underlying infrastructure. It provides orchestration and packaging of several AWS services to ease the deployment process, although it is not based on Kubernetes, which may surprise some users. Harness Platform enhances Fargate by offering flexible pipelines supporting multiple deployment models, including ECS on EC2, EKS, and Fargate, allowing for confident deployments and rollbacks across different environments. Despite its advantages, Fargate can become costly, and understanding its pricing is crucial for managing cloud expenses. AWS offers various container services, including ECS, EKS, and Fargate, each with its own strengths and limitations, providing builders with choices depending on their infrastructure needs. The Harness Platform further extends these capabilities by enabling robust pipelines that support diverse infrastructure models, showcasing its flexibility and adaptability to evolving cloud services and orchestrators.

Tag management in Harness is a feature that enhances Continuous Delivery by allowing developers to organize, search, and filter application components such as services, environments, and pipelines, which aids in better cost allocation and resource management. Tagging is a widely-used method across various software tools, not just in DevOps, for managing data complexity by categorizing it into logical components. As application environments and development teams grow, so does the need for effective tagging to manage complexity, economic constraints, and various cloud strategies. Harness's tagging capabilities enable users to categorize components by team, mark them for reuse, preservation, or deletion, and allocate costs effectively. The tagging feature, which can be configured in the Harness setup, allows users to create and manage tags for supported components, facilitating a better understanding of applications and strategies. Future enhancements are planned to expand the tagging capabilities further, reflecting the dynamic needs of businesses in organizing their development processes.

Decoupling deployment from release enhances software development processes by allowing unfinished code to be deployed without being released to users, thus enabling safe testing in production and improving release safety and velocity. This approach supports progressive delivery and minimizes risks by allowing teams to control feature exposure and perform experiments such as A/B tests, all while monitoring system and user behavior metrics. The strategy helps in preventing the unintended exposure of work-in-progress code and allows developers to safely test new features in production environments by gradually increasing the exposure from internal users to larger user populations. The method of "un-releasing" rather than rolling back changes ensures that teams can address issues without redeploying previous versions, thereby maintaining stability and fostering innovation. Split's Feature Data Platform facilitates this by offering tools to manage feature flags, conduct experiments, and make data-driven decisions, ultimately accelerating the development cycle without compromising safety or quality.

Kubernetes is evolving to handle more complex workloads with a focus on hybrid cloud solutions, customized clusters, and serverless computing, enhancing its flexibility, efficiency, and scalability in modern application deployment. The platform is transitioning from a deployment tool to a customizable framework where application-specific modifications, such as Operators and Custom Resource Definitions (CRDs), are becoming more prevalent. Despite its rapid evolution, Kubernetes has stabilized in recent updates, emphasizing platform maturity and security improvements, although vulnerabilities occasionally arise. The increased adoption of Kubernetes across various infrastructures highlights its expanding role in distributed computing, with solutions like Kubernetes Federation addressing multi-cluster management. As serverless computing gains traction, Kubernetes adapts by integrating projects like KNative, ensuring its continued relevance. The ecosystem's growth is supported by cloud-native projects that align with container orchestration, and while serverless functions introduce competition, Kubernetes remains vital for long-lived workloads. The technology continues to attract investment and development, with organizations embracing it to modernize platforms and enhance operational efficiency.

Decoupling deployment from release is a strategy that allows software teams to deploy code without immediately impacting users, thus enabling safer, more flexible, and gradual rollouts. This method uses feature flags to dynamically control the exposure of new features, supporting internal testing and progressive delivery. The approach helps mitigate risks associated with traditional "big bang" deployments by allowing code to be pushed to production without being activated for all users at once. Historically, the term "release" was used primarily as a noun, but as software delivery practices evolve, it's increasingly used as a verb, reflecting the gradual and controlled nature of modern releases. The Split Feature Data Platform exemplifies this approach by providing tools for feature management, allowing teams to confidently experiment and deploy features without compromising system stability.

Kubernetes is a powerful platform for orchestrating distributed computing but presents challenges when integrating legacy, stateful applications not built with its paradigm in mind. Stateless applications are better suited for Kubernetes, while stateful applications require careful consideration of clustering, replication, and persistence. The platform's evolution has improved support for external storage, but issues like distributed system fallacies and the complexity of containerizing existing applications persist. Modern solutions and tools, such as Apache Kafka and Portworx, address some of these challenges, while capabilities like the Horizontal Pod Autoscaler aid in process scaling. Developers are encouraged to adopt an incremental approach, starting with stateless applications and gradually incorporating more complex workloads, using platforms like Harness to manage deployments across diverse infrastructures.

Harness offers a Continuous Delivery Abstraction Model (CDAM) that simplifies cloud migrations by enabling businesses to switch services, infrastructure, and tools without rewriting deployment pipelines. This model supports various deployment strategies and cloud providers, facilitating efficient migrations such as lift-and-shift, re-platforming, and hybrid cloud architectures. Harness's approach allows for automation and integration, significantly reducing migration time and complexity, as demonstrated by a seamless migration that took only 76 minutes. By providing abstractions between services, infrastructure, and toolchains, Harness minimizes dependencies and improves velocity in deployment processes. The platform supports a wide range of artifacts, provisioners, cloud providers, verifications, secrets management, deployment, and notification strategies, making it possible to switch components without extensive reconfiguration. Since its launch, companies like Citi, SoulCycle, and Advanced Software have accelerated their cloud migrations using Harness CDAM, benefiting from enhanced portability, management, and scaling.