August 2022 Summaries

The latest Gartner Hype Cycle™ for Agile and DevOps 2022 report highlights that feature management has replaced hypothesis-driven development as the dominant approach to experimentation in software engineering. Feature management is poised to enter mass adoption with a high benefit rating, according to Table 1 in the report. This shift from focusing on technical capabilities to potential business impact marks a significant change for many organizations. Gartner emphasizes that feedback loops are crucial for businesses and can be most useful when they're short and targeted. Feature management enables companies to create a culture of experimentation, allowing them to gather timely, fine-grained feedback and data on new features and capabilities. This leads to faster development cycles, reduced technical debt, and more personalized user experiences.

LaunchDarkly's annual conference, Trajectory, is set to take place this year focusing on "Modernizing Software Development." The event aims to bring together customers, team members, and the community to discuss topics related to feature management and software development. This year's themes include Building and Shipping Software, Collaborating Across Teams, Experimentation, Platform Engineering, Release Management, and Lessons Learned. Trajectory offers an opportunity for participants to share their experiences, learn from each other, and explore new ideas in the ever-evolving world of software development. The Call for Papers (CFP) is currently open for talks and workshops related to these themes.

Continuous delivery (CD) and continuous integration (CI) are crucial for optimizing DevOps workflow, shortening development lifecycle, and maintaining high-quality code. Deployment automation tools like Jenkins, AWS CodeDeploy, TeamCity, CircleCI, and GitLab help remove the need for manual deployments and largely eliminate human error. These tools offer various features such as plugins, integration with other platforms, support for multiple OSs, and pricing plans suitable for different team sizes and needs. By implementing these tools, teams can focus on writing code while ensuring frequent creation and release of high-quality software.

This article discusses two ways of representing systems using diagrams: Unified Modeling Language (UML) deployment diagrams and entity-relationship diagrams (ERD). UML deployment diagrams show execution architecture, including nodes such as hardware and software environments, the middleware connecting them, communication links, and file placement. ERDs illustrate relationships of entities stored in a database, representing data storage in the database. Key components of an ERD include entities, actions, attributes, connecting lines, and cardinality. The main difference between UML deployment diagrams and ERDs is that the former models system hardware and its interaction with software, while the latter models databases and their entity relationships.

Cloud computing offers various deployment models to cater to different user needs. Public cloud is managed by a third-party supplier and provides ease of setup, high customer support, and scalability but may have limited customization options and resource limits. Private cloud allows companies to control their own services and tailor solutions to specific needs but requires more maintenance and might limit scalability. Hybrid cloud combines the best features of both public and private clouds, offering adaptability and choice in workload deployment. Community cloud is shared by related organizations for cost savings and collaboration opportunities. Multi-cloud allows businesses to use multiple cloud vendors for flexibility and scalability but may require more complex management. Security risks vary across models, with private clouds generally offering the highest level of security. Costs also differ, with public clouds being the most affordable option and private clouds requiring significant resources and maintenance.

Feature flagging, a method of changing an application without redeployment, has evolved from simple configuration files to complex systems requiring manual intervention. LaunchDarkly is a feature management platform that aims to simplify this process by providing a centralized platform for dynamic feature flags and user targeting. The OpenFeature project, created by Dynatrace and supported by multiple stakeholder companies, aims to establish a standard unified API and SDK for feature flagging across open source and commercial offerings. LaunchDarkly is heavily invested in the OpenFeature project and believes that standardizing data structure and APIs around feature flags will create better integrations and experiences for teams using these features. The standardization will also enable new use cases and operational improvements, benefiting DevOps and SRE teams as well.

Server side experimentation is crucial in agile environments as it allows developers to conduct tests on web servers without impacting users' experience. This type of testing can include A/B testing, multivariate testing, or multi-armed bandit testing. It broadens the testing options and enables deep experimentation with minimal impact on software performance. Server side experimentation is important because it allows organizations to run more sophisticated experiments, including those focused on backend infrastructure. This method also lets teams access and test their backend algorithms, infrastructure, third-party integrations, and databases, providing more data and capabilities for constant product improvement. Feature flags are a key component of server side experimentation as they allow developers to selectively turn features on or off during runtime without needing to push new code or create a new deployment. This enables rapid feature experimentation in a controlled manner. Agile testing practices can be applied with feature flags, allowing teams to work on and release products in short planned sprints. By breaking down releases into smaller pieces and measuring feedback from controlled releases, the server side experimentation cycle can be made more efficient.

Achieving high performance in engineering teams requires more than just tools; it involves a combination of factors like process, consistency, leadership, and culture. The text highlights insights from a webinar featuring Cody De Arkland and Gene Kim, where they discuss how teams can enhance their performance by focusing on key metrics such as lead time, deployment frequency, and change failure rate, as outlined in DORA's assessments. High-performing teams deploy frequently, resolve issues swiftly, and maintain low failure rates, while elite teams surpass these benchmarks with even faster deployments and recovery times. Enhancing team performance involves improving culture, architecture, visibility, and reducing deployment anxiety, often by decoupling releases from deployments using feature flags. Gene Kim emphasizes the importance of managing technical debt and dedicating significant resources to infrastructure and platform development to avoid preordained failure. The journey to becoming a high-performing team is continuous, requiring ongoing commitment to optimization and improvement.

LaunchDarkly offers a variety of SDKs that go beyond just wrapping API calls and provide additional layers of functionality. The SDKs are divided into two categories: client-side and server-side, with mobile SDKs also considered client-side. Client-side SDKs have different security requirements and bandwidth considerations compared to server-side SDKs. LaunchDarkly's SDKs ensure fast flag evaluation by using in-memory caching and streaming updates for both server-side and client-side applications. The SDKs also handle analytics data automatically, reducing load and limiting bandwidth usage. Additionally, the SDKs make it easy to configure additional settings such as persistent data stores and Relay Proxy. While the API can be used for more general actions or automations, the SDK is recommended for most use cases due to its efficiency and built-in features.

LaunchDarkly's survey on psychological safety reveals that developers desire to improve their processes without causing problems. Release assurance, which involves knowing what version of software a user has received and its impact, is crucial for understanding necessary changes. To build release assurance into the software development life cycle, developers should know what they are shipping, collect objective and subjective data about users' experiences, and take action on the information gathered. This approach helps improve product fit and analysis by providing valuable insights into user behavior and preferences.