the machine • No Hypervisor overhead. – Apps directly run on CPU – Rack space & power & more free CPU cycles • Run different versions of OS on same server – Run different OSes on the same server
the machine • No Hypervisor overhead. – Apps directly run on CPU – Rack space & power & more free CPU cycles • Run different versions of OS on same server – Run different OSes on the same server • Process isolaVon & security
“secret sauce” – Everything runs in containers – Not open source • Mesos – Developed by Twiger – Open sourced “IT’S A WAY OF STITCHING TOGETHER A COLLECTION OF MACHINES INTO, BASICALLY, A BIG COMPUTER.”
is hard • It’s hard to glue the technology – LXC + cgroups – lmcky: “Let me contain that for you” • open source version of Google’s container stack • There is no standard image format
deployed by ops teams • Docker might be “the first true DevOps tool” – Increased shipment cadence – Fast moving engineering culture • Sysadmins no more.
deployed by ops teams • Docker might be “the first true DevOps tool” – Increased shipment cadence – Fast moving engineering culture • Sysadmins no more. – A great abstracVon over infrastructure.
Docker containers across nodes – Inspired by Google’s internal systems like Borg/ Omega – Open source – 10+ years of experience in running containers – ContribuVons by Microso[ & IBM etc.
Docker containers across nodes – Inspired by Google’s internal systems like Borg/ Omega – Open source – 10+ years of experience in running containers – ContribuVons by Microso[ & IBM etc. • fleet (by CoreOS) – for clustering CoreOS nodes – Open source
• Inside the Borg and Mesos (Verge) • Google Open Sources Its Secret Weapon in Cloud CompuVng: Kubernetes (Wired) • Docker: hgp://docker.com • CoreOS: hgp://coreos.com • Running Docker on Windows (docker.com) • Docker on Azure (azure.com) • Omega: flexible, scalable schedulers for large compute clusters SIGOPS EuroSys 2013, ACM, pp. 351-‐364 (Google Research)