May 2023 Summaries

Temporal has experienced significant growth since its inception in September 2020. The company now consists of 116 employees across various teams, including Server, SDK, Cloud, Infra, Developer Tools, Security, and a "10x" team. In addition to engineering roles, Temporal has built other critical functions such as Go To Market, Technical Writing, Education, Customer Success, Recruiting, Finance, Product, and Design teams. The company raised its B round in December 2021 and is actively hiring to accelerate investments across multiple areas. Temporal transitioned from a primarily local company to a fully remote one, unlocking access to great hires outside of the Seattle area. However, the company's geographical makeup presents challenges in terms of effective collaboration. Temporal is still figuring out how to manage this situation and would like to learn from other companies that have faced similar issues. The company uses various collaboration tools such as Slack, Zoom, Notion, GitHub Actions, and k9s. Temporal also leverages its own product for building the Control Plane, which manages resources, deployment of software, monitoring, alerting, and handling failures. The architecture of the Control Plane is designed to minimize and contain the impact of any issue within an individual Cell. One of the core characteristics of Temporal is that a Temporal Cluster can scale linearly with the database and compute resources given to it. The company continues to learn the best approach for balancing between scaling out and up, investing in both high-scale individual Cells and Namespaces. For 2023, Temporal plans to focus on people, evolving its work processes, planning, and collaboration methods as the company grows.

Temporal has implemented Just in Time (JIT) access using Common Fate, which limits user access only to the time they need it and for the duration of their request. This approach significantly improves security posture by reducing hours of access from 100% to just 23.8%. By implementing JIT access, Temporal has reduced the number of access hours to under 5% of the previous amount. The use of JIT access also decreases background noise for detection and response tools, making it easier to identify anomalous activity. Additionally, giving users what they want initially can help with a smoother transition to JIT access while still providing significant security benefits.

In the May newsletter, Temporal announced an extended deadline for CFP submissions to Replay until May 31st. They also shared updates on technology, including the release of Schedules to Public Preview and new features in the Cloud like Audit Logging and tcld API Key management. The .NET SDK feature is now complete, and Temporal has expanded its services with scale tests, production readiness, technical onboarding, and design sessions. Upcoming events include Money 20/20, Domain Driven Design, JCon, and Replay conference in 2023. Community favorites include the introduction of Temporal .NET SDK, compensating actions article, and using the Temporal CLI blog.

The first version of the Temporal Visual Studio Code extension has been released, initially focusing on improving TypeScript workflow debugging experiences. It allows users to load Event Histories by Workflow Id or JSON file, easily start a replay, and set breakpoints on events rather than just code. A video tutorial is available for learning how to use the extension. The Temporal Server connects by default to localhost:7233, but this can be configured to connect to different servers. Contributions and bug reports are welcome.

The Saga Design Pattern is useful when your logic involves multiple steps across various services or databases, and partial execution is undesirable. It functions as a state machine tracking program progress, preventing multiple credit card charges, reverting if necessary, and ensuring consistent recovery in case of failure. A common example is trip planning, where booking an airplane ticket, reserving a hotel, and getting a guided tour ticket are all coupled. The Saga Pattern compensates for failures by undoing actions when needed. It consists of two parts: defined behavior for "going backwards" (compensations) and behavior for striving towards forward progress (saving state). Implementing sagas in code can be simplified using Temporal, which automatically saves state and retries on failure at any level.

Scaling a Temporal Cluster involves addressing various workload patterns and operational goals, and this guide simplifies the process by focusing on key metrics and terminology. While using Temporal Cloud offers an easy scaling solution, the guide details a step-by-step approach for scaling a self-hosted Temporal Cluster on Kubernetes. This involves iterative cycles of loading, measuring, and scaling to transition from development to production-level configurations, with attention to aspects like Kubernetes resource management, shard count configuration, and polling optimization. The process includes adjusting the number of shards to reduce lock contention, optimizing CPU and memory usage, and improving poll sync rates to enhance throughput. The guide also emphasizes the importance of maintaining Service Level Objectives (SLOs) for request latency and Schedule-to-Start latency, using Prometheus and Grafana for monitoring. Through these adjustments, the cluster's performance significantly improves, achieving a substantial increase in State Transitions per second while maintaining database efficiency.

The newsletter from May 23rd highlights several important updates and events related to Temporal, a platform for managing software applications. Notably, the deadline for submitting talks to the Replay conference has been extended to May 31st, and the public preview of Schedules is set to launch next month for Cloud customers. New features such as Audit Logging and API Key management are also being introduced, with some available for public preview and others in private preview. The .NET SDK feature is now complete, expanding Temporal's supported languages, and the Replay conference is set to offer a wide range of talks and workshops, including Temporal trainings in Go, Java, and Typescript. The newsletter also promotes upcoming events like Money 20/20 and Domain Driven Design in Amsterdam, as well as JCon in Cologne, with opportunities for community engagement and ticket giveaways. Additionally, an advanced course, Temporal 102 with Go, has been launched to help users further develop their skills in testing, debugging, deploying, and updating applications.

The Replay Conference has extended its Call for Papers (CFP) deadline until May 31st to give more people the opportunity to share their ideas, especially since this is the first year of the new format. They are seeking talks on various topics such as backend engineering platforms and technologies, innovative general engineering practices, Temporal use cases, and improving complexity and reliability in services and systems. Additionally, a speaker program has been introduced to assist individuals with refining their talk presentations and public speaking skills. Conference tickets are available for purchase, and the event will take place in Seattle during summer.

Temporal Cloud has launched a public preview of its Audit Logging feature, which allows users to keep track of actions performed in the platform. This feature is crucial for maintaining transparency, security, and accountability within organizations. The preview release logs information for administrative operations and global configuration changes. Future updates will include support for additional destination sinks and new auditable events. General availability is expected within a quarter.

Temporal 102 with Go`, a new training course from the creators of Temporal, is now generally available and free online. The course builds on `Temporal 101 with Go` by focusing on best practices and key concepts for deploying and maintaining applications in production, rather than covering additional features. This course was developed based on feedback from users, including those who participated in an early access beta version, as well as interviews with experts within the company. Developers who have taken `Temporal 101` with Go or basic proficiency in another language SDK may benefit from this course, which will be ported to other SDKs in the future.

The compensating action pattern is a design pattern for handling failure amongst distributed but related services. It provides transaction-like guarantees for a sequence of operations in distributed systems by either letting the sequence run successfully to completion or undoing state changes made by executed operations if there's a failure. This sequence of steps is called a "long-running transaction". Compensating actions are an important component of the saga design pattern, which ensures that there is a way to "go backwards" (undo) and end up with a consistent state in case of failure.