Jeff Dean’s “Designs, Lessons and Advice from Building Large Distributed Systems” is essential reading for anyone building or operating distributed infrastructure. This talk, presented at multiple conferences including LADIS and various tech events, distills decades of experience at Google and now at Google Research. If you’re designing systems that scale beyond a single machine, you…
MapReduce remains a foundational programming model for processing large-scale distributed datasets, though its role has evolved alongside modern data processing frameworks. Hadoop implementations and other MapReduce systems require objective performance evaluation to guide infrastructure decisions and optimization efforts. PUMA (Princeton University MapReduce Benchmark) is a comprehensive benchmark suite developed by Faraz Ahmad and colleagues to…
TeraSort is Hadoop’s standard benchmark for measuring distributed sorting performance across clusters. It consists of three components: TeraGen generates random input data, TeraSort performs the distributed sort, and TeraValidate confirms the output is globally sorted. Modern Hadoop deployments typically run on Hadoop 3.x or higher, with improvements in YARN resource management, better performance tuning options,…
Modern datacenters rely on distributed storage and processing systems designed to handle massive datasets across clusters of commodity hardware. Understanding these systems is essential for anyone working with infrastructure at scale. Here’s an overview of the major architectures and implementations you’ll encounter. Storage Systems Google File System (GFS) and successors GFS established the foundation for…
Cosmos is Microsoft’s internal petabyte-scale distributed data storage and query system, built to handle massive analytical workloads across their infrastructure. While Microsoft hasn’t published formal technical papers on Cosmos, information has surfaced through conference talks and engineering blogs over the years. Architecture and Scale Cosmos operates at truly massive scale: Storage capacity: Approximately 62 physical…
Colossus is Google’s successor to the Google File System (GFS), designed to handle the massive scale of data storage across Google’s infrastructure. While GFS was documented in detail in the 2003 SOSP paper, Colossus information remains sparse in academic literature—much of what we know comes from talks and interviews with Google engineers. Colossus serves as…
Historical Archive: This post lists conferences from 2012. The submission deadlines and venues are no longer current. For up-to-date conference information, check the official conference websites or resources like Conference Alert and CCF Deadlines. I am trying to find out the top conferences that have the largest average number of citations in the last 5…
Hadoop 1.x reached end-of-life in 2014 and is no longer maintained. This post documents a deprecated architecture for historical reference only. For new deployments, use Hadoop 3.x+ with YARN, which offers significant improvements in resource management, multi-tenancy, and reliability. See the Apache Hadoop documentation for current versions. For managed services, consider AWS EMR, Google Dataproc,…
Understanding distributed systems and cloud computing starts with the literature. These papers form the foundation for anyone working in the space — they’re not just historical artifacts, they’re still actively referenced and their design principles remain relevant. Internet-Scale Systems and Datacenters The Google cluster architecture paper remains essential reading. It laid out the practical realities…
Historical Archive: This post lists conferences from 2012. The submission deadlines and venues are no longer current. For up-to-date conference information, check the official conference websites or resources like Conference Alert and CCF Deadlines. This post lists important conferences related to Cloud Computing in year 2013. SOSP 2013 SOSP’13: The 24th ACM Symposium on Operating…
Gitosis is unmaintained and shouldn’t be used for new deployments. Use Gitolite or Forgejo instead. This guide covers Gitolite, which provides fine-grained access control, is actively maintained, and handles multiple repositories under a single system user. Prerequisites A server running Linux (Debian, Ubuntu, CentOS, or similar) SSH access to the server with sudo privileges A…
Running benchmarks after a Hadoop installation helps validate that your cluster works correctly and gives you baseline performance metrics. The Sort benchmark is a standard test that exercises the MapReduce execution engine and HDFS, making it useful for stress-testing both the distributed storage layer and computation framework. What the Sort Benchmark Does The Sort benchmark…
Historical Archive: This post lists conferences from 2011. The submission deadlines and venues are no longer current. For up-to-date conference information, check the official conference websites or resources like Conference Alert and CCF Deadlines. This post lists important conferences on Cloud Computing in year 2012. OSDI 2012 10th USENIX Symposium on Operating Systems Design and…
Archived Content (2010): This post describes a research project from 2010. The tools and versions mentioned (Hadoop 0.20, MapReduce Streaming) are historically significant but have been largely superseded by modern distributed build systems like Bazel Remote Execution and cloud-native CI/CD pipelines. mrcc – A Distributed C Compiler System on MapReduce Original Project Date: January 2010…