Rootconf 2014

On devops and cloud infrastructure

Scaling to 2000 request per second with MongoDB

Submitted by Suchit Puri (@suchitpuri) on Wednesday, 8 January 2014

Section: Crisp talk Technical level: Intermediate Status: Confirmed & Scheduled

View proposal in schedule


The Objective of this session is to showcase my experiences with developing , scalling to 2000 req/sec and performance testing the website for one of the biggest media houses of the UK .


In this session i will talk about

1) How important is "Schema Design" when scalling mongo.
2) Mongo Sharding .
3) Locking mechanisms in Mongo ( DB level locks , mongo instance level locking ).
4) Replication mechanisms in a mongo shard.
5) Performance testing the applicatoin , we used
6) The Good side of mongo - the flixibility it provides to your application
7) Problems faced with migrations.

Speaker bio

I am an agilist and ardent follower of XP best practices like TDD, Pair Programming, Continuous Integration and continuous deployments. I have worked for clients ranging from small startups to big MNC's ; getting involved during project inceptions/client management, juggling many hats; doing analysis, coding, deployments and automation during a project's life-cycle. I have also been doing consulting on Agile Adoption , TDD ,CI.
I work for Thoughtworks.




  •   Ajey Gore (@ajeygore) 5 years ago


    I would like to see the reasons for mongo to scale 2000 requests per second, I think few things which makes mongo very different is the way it deals with data and structures all slaves and masters! will you explain about embedding documents and when to embed (this is usually done to avoid unnecessary join calls and other db calls), when to use mongos and when to use sharding vs replication or everything?


  •   Suchit Puri (@suchitpuri) Proposer 5 years ago

    I will be covering the Schema Design, when to go with Embedded Vs Reference. With some considerations like the document size , joins .. etc
    I am planning to cover Replication how its done using Oplog capped collection.
    I am planning to cover how to scale horizontally by using sharding , and how important is key on which it is sharded to avoid skewed shards.

    please let me know if you have more questions.

    Suchit Puri

  •   Sreekandh Balakrishnan (@gnuyoga) 5 years ago

    Summarizing the call.

    What problems are we attending
    1. Migration scirpts
    2. Testable
    3. Scalable
    4. Distributed

    What are the +ive and -ive
    How did the adoption of the tool impacted the team
    SQL Vs NoSQL. When to use what
    Option in NoSQL ( Mongo, Couch, Dynamo, Cassadra)
    Performance benefits
    Atomicity of transaction
    Eventual consistency
    Support available for NoSQL is it important ?

    Share your experience rather than what;s said in public domain. Get a outline and let me know we can review it again together.

  •   Suchit Puri (@suchitpuri) Proposer 5 years ago

    Hi Sreekanth,

    Have uploaded a very raw draft of the presentation , please check it out

    i need to refine the presentation more , but it covers the basic outline of the presentation.

    Let me know when can we catchup and discuss the same.


  •   Suchit Puri (@suchitpuri) Proposer 5 years ago
  •   Suchit Puri (@suchitpuri) Proposer 5 years ago (edited 5 years ago)

    Please see the new slides at , as requested by you to fit in the theme.

Login with Twitter or Google to leave a comment