Top menu

Mesos Cluster Management Training

 

Introduction

Apache Mesos is a cluster manager that provides efficient resource isolation and sharing across distributed applications, or frameworks. Think of it as the “kernel” for your data center.

Features:

  • Fault-tolerant replicated master using ZooKeeper

  • Scalability to 10,000s of nodes

  • Isolation between tasks with Linux Containers

  • Multi-resource scheduling (memory and CPU aware)

  • Java, Python and C++ APIs for developing new parallel applications

  • Web UI for viewing cluster state

Apache Mesos – Intentions

  • Avoid multiple specialised clusters
  • Avoid the need for “silo’ed” clusters
  • Avoid the lower utilization that this brings
  • Offer the ability for multi tenancy
  • Which offers  Higher utilisation,  Scalability,  Lower hardware / maintenance costs, -Higher fault tolerance
  • Program for the Data Center
mesos training

Course Outline

1. Introduction to Mesos

  • How to Use Mesos
  • Mesos as a Deployment System
  • Mesos as an Execution Platform

 

2. Getting Started with Mesos

  • Frameworks
  • Masters and Slaves
    • The Masters
    • The Slaves
  • Resources
    • Configuring Custom Resources
    • Configuring Slave Attributes
  • Roles
    • Static and Dynamic Slave Reservations
  • Tasks and Executors
    • CommandExecutor
  • Understanding mesos.proto
  • Not Managed by Mesos

 

3. Porting an Existing Application to Mesos

  • Moving a Web Application to Mesos
  • Setting Up Marathon
  • Using Marathon
    • Scaling Your Application
    • Using Placement Constraints
    • Running Dockerized Applications
    • Health Checks
    • Application Versioning and Rolling Upgrades
    • The Event Bus
    • Setting Up HAProxy with Marathon
  • Running Mesos Frameworks on Marathon
    • What Is Chronos?
    • Running Chronos on Marathon
    • Chronos Operational Concerns
    • Chronos on Marathon
  • Alternatives to Marathon + Chronos
    • Singularity
    • Aurora

 

4. Creating a New Framework for Mesos

  • The Scheduler
    • Pool of Servers Scheduler
    • Work Queue Scheduler
    • Job Processor Scheduler
  • Useless Remote BASH
  • Implementing a Basic Job Processor
  • Matching Tasks to Offers
    • Bridging the Semantic Gap Between Offers and Jobs
  • Adding High Availability
  • Adding Reconciliation
  • Advanced Scheduler Techniques
    • Distributed Communication
    • Forced Failover
    • Consolidating Offers
    • Hardening Your Scheduler
    • Framework UI
    • Allocating Ports
    • Checkpointing
    • CommandInfo

 

5. Building a Mesos Executor

  • The Executor
    • Building a Work Queue’s Worker
    • Running Pickled Tasks
    • Sharing Resources
    • Better Babysitting
    • Augmented Logging
  • Rewriting the CommandExecutor
  • Bootstrapping Executor Installation
  • Adding Heartbeats
  • Advanced Executor Features
    • Progress Reporting
    • Adding Remote Logging
    • Multiple Tasks

 

6. Advanced Topics in Mesos

  • Libprocess and the Actor Model
  • The Consistency Model
    • How Is Slave Failure Handled?
    • How Is Master Failure Handled? (Or, the Registry)
    • Reconciliation During Failover
  • Containerizers
    • Using Docker
  • The New Offer API
    • Framework Dynamic Reservations API
    • Persistent Volumes for Databases

 

7. The Future of Mesos

  • Multitenant Workloads
  • Oversubscription
  • Databases and Turnkey Infrastructure
  • IP per Container

 

 

 

info@bigdatatraining.in

http://www.bigdatatraining.in/contact/

Call – +91 97899 68765 / 044 – 42645495

Big Data Training Bangalore Hadoop Training in Bangalore, 2013