Upgrade to Pro — share decks privately, control downloads, hide ads and more …

MongoDB Replica Sets on Windows Azure

Sridhar Nanjundeswaran
February 24, 2012

MongoDB Replica Sets on Windows Azure

Presented at MongoDB Silicon Valley in Dec 2011

Sridhar Nanjundeswaran

February 24, 2012


  1. Deploying MongoDB on Azure Eduard Koller, Program Manager, Microsoft Sridhar

    Nanjundeswaran, Software Engineer, 10gen David Makogon, Cloud Solution Architect, Microsoft
  2. Windows Azure Platform • Microsoft’s cloud platform to build, host

    and scale applications • Internet-scale computing and services • Hosted in Microsoft data centers Compute Storage Management CDN
  3. Why Windows Azure? • More time coding, less time managing

    hardware and/or deploying applications • Scale your applications up or out on demand • Trust: Enterprise class service backed by reliable SLAs • Efficiency: Microsoft operates the Windows Azure platform at internet-scale
  4. Azure Interoperability • Data portability • Cloud Standards • Ease

    of Migration and Deployment • Developer choice
  5. Why MongoDB on Azure? • High availability built in •

    Data replication and durability built in • You can scale up on demand your application and the MongoDB database • You can scale out on demand your application and the MongoDB database(1)
  6. MongoDB on Azure implementation history • Manual setup of replica

    sets – demo’ed at MongoSV 2010. David Makogon • Single instance with standby – demo’ed at MongoSF 2011. Sridhar Nanjundeswaran • Replica set support - now
  7. Highlights • Uses standard MongoDB binaries • Code is open

    sourced • Works with VS Express • Automatic replica set initiation on deploy • Survives reboots of instances • Integration with Azure diagnostics • Data persisted on blob storage
  8. Challenges • Debugging when building – On emulator debug –

    On deployment, trace + tools such as Cerebrata • IP potentially changes on reboot – DNS within mongodb. New cloud command • Bandwidth – Requires at least medium instance – Don’t store mongod.log on blob storage • Testing IP change – Reduce instance count and increase
  9. The Solution • Source – https://github.com/mongodb/mongo-azure • Documentation – http://www.mongodb.org/display/DOCS/Mon

    goDB+on+Azure • Issues – mongodb-user google group – #mongodb IRC – https://jira.mongodb.org/browse/AZURE
  10. The Future • Scale out using replica sets • MongoDB

    Monitoring • Backup and Recovery • Sharding