YARN Forum

YARN vs. Mesos

  • #33347
    Grant Priestley

    Hi All,

    I am trying to find more information regarding the key differentiators between YARN and Mesos, in terms of resource provisioning (eg. does YARN support resource provisioning of non-JAVA or MapReduce jobs). There is some information scattered across the internet, however no “official” documentation / comparison between the tools.


to create new topics or reply. | New User Registration

  • Author
  • #33350
    Grant Priestley

    Hi Priyank,

    Yes I have read the post you have referred to; however it is quite dated (sept 2011). There have been significant advances in YARN since then Eg. CPU allocation using cgroups.


    Arun Murthy

    Grant, the quora post is reasonably accurate about high-level differences such as YARN has resource-requests, while Mesos has offer/reject with concept of resource-requests etc.

    Since then, like you noted, YARN has made significant strides with expanded features (multi-resource scheduling, cgroups, significant number of applications such as Storm, Giraph, HOYA etc.) and has matured considerably. For e.g. Yahoo has been running 35000 nodes of YARN in production for over 8 months now. Some more details: http://strata.oreilly.com/2013/06/moving-from-batch-to-continuous-computing-at-yahoo.html. So, the take away is primarily around its’ maturity and battle-hardness since that post.

    W.r.t Mesos and it’s maturity, deployment-scale etc., I don’t think I’m the right person to answer it since I don’t want to mis-represent them. You might want to check with the Mesos community directly. Thanks.

    Grant Priestley


    Thank you for your reply. We are about to commence some comparison testing of mesos against yarn and was wondering if there are any benchmarks that are available for Yarn? I’m currently discuss the possible release of the comparison to a Hadoop / Big Data community for everyone to share.


You must be to reply to this topic. | Create Account

Support from the Experts

A HDP Support Subscription connects you experts with deep experience running Apache Hadoop in production, at-scale on the most demanding workloads.

Enterprise Support »

Become HDP Certified

Real world training designed by the core architects of Hadoop. Scenario-based training courses are available in-classroom or online from anywhere in the world

Training »

Hortonworks Data Platform
The Hortonworks Data Platform is a 100% open source distribution of Apache Hadoop that is truly enterprise grade having been built, tested and hardened with enterprise rigor.
Get started with Sandbox
Hortonworks Sandbox is a self-contained virtual machine with Apache Hadoop pre-configured alongside a set of hands-on, step-by-step Hadoop tutorials.
Modern Data Architecture
Tackle the challenges of big data. Hadoop integrates with existing EDW, RDBMS and MPP systems to deliver lower cost, higher capacity infrastructure.