BOSS: Automatically Identifying Performance Bottlenecks through Big Data

    January 11, 2017

    Introduction As the centralized performance team of LinkedIn, our mission is to make LinkedIn pages load faster. We help each engineering team try to hit their page load time goals through various optimization efforts. One common question we need to answer when trying to decrease page load time is: where is the performance bottleneck? In other words, where...

  • async21

    Asynchronous Processing and Multithreading in Apache Samza, Part II: Experiments and Evaluation

    January 6, 2017

    This post is the second in a series discussing asynchronous processing and multithreading in Apache Samza. In the previous post, we explored the design and architecture of the new AsyncStreamTask API and the asynchronous event loop. In this post, we will focus on the study of the performance of this feature with benchmark Samza jobs. Some of the interesting...

  • Async1

    Asynchronous Processing and Multithreading in Apache Samza, Part I: Design and Architecture

    January 4, 2017

    As part of the Apache Samza 0.11 release, we rebuilt Samza’s underlying event processing engine to use an asynchronous and parallel processing model. The new model is unique among current open source stream processors because it not only supports running traditional synchronous processing in parallel on multiple threads, but also provides first-class support for...

  • Val1

    Getting to Know Val Markovic

    December 21, 2016

    LinkedIn wouldn't be the company it is today without the engineers who built it. We have no shortage of talented individuals in...

  • Everyday1

    You’re Only as Good as Your Lieutenants

    December 19, 2016

    This post is part of the series “Every Day Is Monday in Operations.” Throughout this series we discuss our challenges, share our war...

  • Inception1

    Inception: How LinkedIn Deals with Exception Logs

    December 16, 2016

    Coauthors: Toon Sripatanaskul and Zhengyu Cai   In early 2012, the LinkedIn Performance team was trying to build a tool to validate...