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

Atlas, a PaaS with batteries included

Atlas, a PaaS with batteries included

a brief description on how the SRE model is applied in Medallia and what is Atlas and why is so important for the rest of the teams

Sebastian Montini

October 12, 2017
Tweet

More Decks by Sebastian Montini

Other Decks in Technology

Transcript

  1. Medallia © Copyright 2017. Confidential. 5 @sebamontini SRE Vision @

    Medallia: No technical debt Picture a world where everything just works.
  2. Medallia © Copyright 2017. Confidential. 6 @sebamontini SRE Vision: No

    technical debt Picture a world where everything just works.
  3. Medallia © Copyright 2017. Confidential. 7 @sebamontini Front Line Support

    [Level 1] Production Services [Level 2] Product SREs Developer-SREs SRE Model @ Medallia SREs that work closely with Development to implement SRE Requirements in products SREs that own key initiatives, tools and applications: Telemetry, Configuration Consistency, Sandboxes, etc. 24x7 Routine Run-the-Business Continuous Service Delivery (Change, Incident, Release)
  4. Medallia © Copyright 2017. Confidential. 8 @sebamontini Front Line Support

    [Level 1] Production Services [Level 2] Product SREs Developer-SREs SRE Model @ Medallia SREs that work closely with Development to implement SRE Requirements in products SREs that own key initiatives, tools and applications: Telemetry, Configuration Consistency, Sandboxes, etc. 24x7 Routine Run-the-Business Continuous Service Delivery (Change, Incident, Release) Detect Issues Fix Problems Independently Fix Problems with Owners
  5. Medallia © Copyright 2017. Confidential. 10 @sebamontini • Scalable multi-tenant

    Infrastructure • Single consistent layer for application development “build once run everywhere” • Core Services for microservices • Cost effective redundant commodity hardware • Datacenter management efficiency through automation Why Atlas? Next Generation Cloud Platform
  6. Medallia © Copyright 2017. Confidential. 12 @sebamontini • 6 Datacenters

    (and 2 AWS regions) • 12 racks, 240 compute nodes per DC • 11.000 cores • 80 Tb of RAM • 1 Pb of storage (+300 Tb local storage) • +7.000 tasks • 100% inmutable infrastructure Atlas in a nutshell
  7. AMI Server Terraform + Packer DCIB Docker + Mesos +

    Aurora Microdeployer Titan EC2 AWS Atlas Plenum EBS/S3 Ceph Quagga/OSPF Config Service ELK Msecret ngMonitoring Convoy LBaaS