@andypipe Scaling Twitter Infrastructure with Mesos Chris Aniszczyk (@cra)

Fail Whale and Growth

300M+ 500M+ 80% Active users Tweets per day of users are outside the US 2006 2015 100TB compressed data per Twitter Scale

1 2 3 4 5 6 7 8 9 0% 11% 22% 33% 0% 11% 22% 33% 0% 11% 22% 33% Mixed Utilization Devs requesting CPUs slows you down Static Partitioned Datacenters

0% 11% 22% 33% 0% 11% 22% 33% 0% 11% 22% 33% 0% 33.333% 66.667% 100% Didn’t we learn as children that we should share :)? Resource Sharing!?

• Google was ahead of the game of managing warehouse scale cluster computing: http://youtube.com/watch? v=0ZFMlO98Jkc • NEW (2015): Google Borg paper (https:// research.google.com/pubs/ pub43438.html) • 2009: Universities (Berkeley) was doing research in this area, we decided to partner and hire researchers: https:// amplab.cs.berkeley.edu/tag/mesos/ Evaluated industry + academia

• May 2012: Took university research and spun into a independent open source project at the Apache Foundation: https://blog.twitter.com/2012/ incubating-apache-mesos • June 2013: https://twitter.com/ ApacheMesos/statuses/ 360039441500340224 (graduated) Enter Apache Mesos

Where does Mesos Fit?

build and run a PaaS on top of Mesos: Apache Aurora / Marathon use OpenStack or EC2 or physical machines to run Mesos Where does Mesos Fit?

Mesos treats your datacenter nodes as one big computer! Not as individual machines (or VMs), but as computational resources like CPU cores, memory, disks, network ports, etc… Mesos: Datacenters as 1 CPU

http://mesos.apache.org/documentation/latest/mesos-frameworks/ Frameworks are responsible for using resource offers from Mesos Mesos consists of Frameworks

http://mesos.apache.org/documentation/latest/mesos-frameworks/ Ecosystem of Frameworks

offer hostname 4 CPUs offer hostname 4 CPUs 4 GB RAM offer hostname 4 CPUs 4 GB RAM offer hostname 4 CPUs 4 GB RAM Schedulers get resource offers Executors run tasks Mesos Framework = Scheduler + Executor Mesos at a High Level

Mesos Slave Hadoop task-tracker Mesos Executor Task #1 Task #2 ./ruby XYZ Mesos Slave Docker Executor Docker Executor java -jar XYZ.jar ./xyz Mesos Master Mesos Master Mesos Master Hadoop scheduler Marathon scheduler Zookeeper quorum Master schedules tasks to run on slaves’ available resources; slaves use executors to coordinate execution of tasks Tasks are the unit of execution

Mesos provides fine-grained resource Compute Node Mesos Slave Process Hadoop task-tracker Mesos Executor Task #1 Task #2 ruby XYZ Container (Cgroups) Executor Slaves isolate executors and tasks via containers (dotted line) Compute Node Mesos Slave Process Hadoop task-tracker Task #1 Task #2 Container (Cgroups) Task #3 Containers can GROW AND SHRINK as tasks

http://mesos.apache.org/documentation/latest/powered-by-mesos/ https://mesosphere.com/blog/2015/04/23/apple-details-j-a-r-v-i-s-the-mesos-framework-that-runs-siri/ #PoweredByMesos

http://mesos.apache.org http://www.meetup.com/Austin-Mesos-Users-Group/ Open Source Community

http://mesosphere.io/ Commercial Support

http://mesoscon.org MesosCon 2015!

Thank you for listening! Chris Aniszczyk (@cra) zx@twitter.com http://mesos.apache.org email: {user,dev}@mesos.apache.org https://mesosphere.com/learn/