Rootconf Mini 2024

Abstract

Automation and modernization are foundational to support large infrastructure like that of LinkedIn. Managing a fleet of half a million servers(~500K) across our private data centers at LinkedIn is no small feat. This immense scale demands infrastructure solutions that not only expand capacity but also ensure performance, reliability, and efficiency as we grow. Building for scale involves creating systems that can handle increased loads, adapt to changing demands, and operate seamlessly.

Design patterns are critical in this process. They provide reusable solutions to common architectural challenges, helping us structure components and manage interactions effectively. However, with numerous patterns to choose from, finding the right one for specific contexts can be complex and overwhelming. At LinkedIn, we’ve found the Producer-Consumer pattern to be exceptionally effective. This pattern has been successfully implemented in several of our core infrastructure systems, including the distributed server query system, server console monitoring, and network security monitoring. In this process, we have identified and built general solutions that are repeatable in similar environments, greatly improving engineering efficiency by leveraging proven methodologies.

This talk is a sneak peek into how LinkedIn is operating at scale from an infrastructure point of view. We would like to share how the Producer Consumer Pattern was leveraged in designing 3 of our infrastructure systems and share valuable lessons learned while building several large-scale infrastructure solutions, from choosing the design pattern to selecting the right technology at inception. We will talk about the technological choices and trade-offs involved, and the insights gained from scaling our infrastructure.

Key Takeaways

This talk will walk the audience through 3 instances of the scalable architectures of similar themes (not the same) but solving 3 different problems; and why each one’s choices are different in their own ways. The main takeaways from this talk will be,

• Get introduced to a class of scalable architectures that power Linkedin’s infrastructure solutions.
• Understand how domain requirements and infrastructure restrictions find their way into defining the architectural characteristics
• How can your cognitive biases toward obvious technology choices be tricky?
• How to pick the tech that fits the architectural characteristics at the right inception level required to solve the problem.

Audience

• Technically curious: Anyone, including beginners, interested in learning about highly scalable architectures.
• Architects, Platform Engineers: All experienced engineers seeking to enhance their system design skills and gain a better understanding of how requirements could translate into design choices.