a journey to modern apps with containers, microservices and big data

37
© 2016 Mesosphere, Inc. All Rights Reserved. Edward Hsu, VP of Enterprise DC/OS @edwardihsu JOURNEY TO THE MODERN APP WITH CONTAINERS, MICROSERVICES AND BIG DATA Datacenter Operating System (DC/OS) The simplest way to build, run, scale and manage modern enterprise apps

Upload: edward-hsu

Post on 16-Apr-2017

215 views

Category:

Technology


0 download

TRANSCRIPT

© 2016 Mesosphere, Inc. All Rights Reserved.

Edward Hsu, VP of Enterprise DC/OS@edwardihsu

JOURNEY TO THE MODERN APP WITH CONTAINERS, MICROSERVICES AND BIG DATA

Datacenter Operating System (DC/OS)The simplest way to build, run, scale and manage modern enterprise apps

© 2016 Mesosphere, Inc. All Rights Reserved.

© 2016 Mesosphere, Inc. All Rights Reserved.

INDUSTRY TRANSITIONS

Customer analytics2000s

Electronic customer records1980s

Online customer engagement1990s

Real-time & predictive customer engagement2013+

© 2016 Mesosphere, Inc. All Rights Reserved.

© 2016 Mesosphere, Inc. All Rights Reserved.

© 2016 Mesosphere, Inc. All Rights Reserved.

MAYO CLINIC REVOLUTIONIZING HEALTHCARE

Individualized care recommendations based on 5 million patient records

Key technologies:

● Real-time data ingestion● Distributed search engine● Large-scale processing● Distributed database

© 2016 Mesosphere, Inc. All Rights Reserved.

UBER DISRUPTING MULTIPLE INDUSTRIES

Operations in over 300 cities around the world

Key technologies:

● Distributed logging ● Large-scale data processing● Distributed datastore

© 2016 Mesosphere, Inc. All Rights Reserved.

ARCHITECTURAL SHIFT

TRADITIONAL ENTERPRISE APPLICATIONS

MODERN ENTERPRISE APPLICATIONS

Data

Code

Latency

Users

2+ Billion smartphone users

3+ Billion internet users

Enterprise data growth (CAGR): 40%+

Source: KPCB Internet Trends 2016, EMC Digital Universe 2014

© 2016 Mesosphere, Inc. All Rights Reserved.

ARCHITECTING FOR SCALE AT TWITTER

BEFORE AFTER

© 2016 Mesosphere, Inc. All Rights Reserved.

Streaming

10

THE MODERN ENTERPRISE APP

BIG DATA SERVICESMICROSERVICES

STATELESS CONTAINERS ANALYTICS

Batch Machine Learning

Search

DATABASES

Time Series

SQL / NoSQL

© 2016 Mesosphere, Inc. All Rights Reserved.

MODERN ENTERPRISE APPS

Containers Data Services

MicroservicesOpen source & Social coding1

2

3

4

© 2016 Mesosphere, Inc. All Rights Reserved.

OPEN SOURCE SOFTWARE PROVIDES LEVERAGE1

© 2016 Mesosphere, Inc. All Rights Reserved.

CONTAINERS SIMPLIFY PACKAGING AND DEPLOYMENT

Private Copy

Shared

App 3

Libraries

Guest Operating

System

Infrastructure

App 1

Libs

Host Operating System

Start / stop time One minute to few seconds Milliseconds

Workload density 10 - 100x1x

App 2

Libraries

Guest Operating

System

App 1

Libraries

Guest Operating

SystemApp 2

Libs

App 3

Libs

Virtual Machines Containers

Host Operating System

Hypervisor

Infrastructure

App 4

Libs

App 5

Libs

2

© 2016 Mesosphere, Inc. All Rights Reserved.

MICROSERVICES ENABLE FAST AND FREQUENT RELEASES

Traditional Architecture

Many functions in a single process

Siloed teams

REST

API

s

Microservices Architecture

Cross-functional teams organized around capabilities

Scales individually

Each element of functionality defined as “microservices”

Scales monolithically

3

© 2016 Mesosphere, Inc. All Rights Reserved.

DATA SERVICES PROVIDE CONNECTION AND PERSISTENCE

Data Ingestion

Response

Devices

ClientSensors

Event Bus

Reactive App

Database

Analytics

Use Case Examples

● Anomaly detection

● Personalization

● IoT Applications

● Predictive Analytics

● Machine Learning

4

© 2016 Mesosphere, Inc. All Rights Reserved. 17

BUT WHAT ABOUT THE INFRASTRUCTURE?

CaaSPaaS

Container AppContainer App

Stateful Service #1

V1

Big Data Analytics #1

V1

Stateful Service #2

V1

Stateful Service #2

V2

PaaS

Container App

Team A

Big Data Analytics #1

V2

Big Data Analytics

#2V1

Team B

© 2016 Mesosphere, Inc. All Rights Reserved. 18

BUT WHAT ABOUT THE INFRASTRUCTURE?

CaaSPaaS

Container AppContainer App

Stateful Service #1

V1

Big Data Analytics #1

V1

Stateful Service #2

V1

Stateful Service #2

V2

PaaS

Container App

Team A

Big Data Analytics #1

V2

Big Data Analytics

#2V1

Team B

● Extremely low utilization - less than 10%, due to static partitioning

● Manual operations - Weeks to provision, wasted time managing individual machines

● Difficult to experiment with new tech - months to provision environments to evaluate new technologies

● High risk - Failure from manual operations and difficulties maintaining high availability

© 2015 Mesosphere, Inc. All Rights Reserved.

WHERE HAVE WE SEEN THIS PROBLEM BEFORE?

You are launching Google Chrome.Which core would you like to use?

Core 1 Core 2

Core 3 Core 4

You are launching a distributed app requiring 500 cores.Which of your 50,000 cores would you like to use?

Core 1 Core 2 Core 3 Core 4 Core 5

Core 6 Core 8 Core 9 Core 10Core 7

Core 11 Core 13 Core 14 Core 15Core 12

Core 16 Core 18 Core 19 Core 20Core 17

Core 21 Core 23 Core 24 Core 25Core 22

Core 26 Core 28 Core 29 Core 30Core 27

Core 31 Core 33 Core 34 Core 35Core 32

Core 36 Core 38 Core 39 Core 40Core 37

Core 41 Core 43 Core 44 Core 45Core 42

Core 46 Core 48 Core 49 Core 50Core 47

© 2015 Mesosphere, Inc. All Rights Reserved.

MODERN ENTERPRISE APPS REQUIRE WE OPERATE AT THE LOGICAL DATACENTER LEVEL

Traditional App Modern Enterprise App

© 2015 Mesosphere, Inc. All Rights Reserved.

YES, THIS TOO CAN BE SOLVED WITH ABSTRACTION

Scheduler

Executor

Task Launch

Task Launch

Task Status

Task Status

Resource Offer

ExecutorExecutorExecutor

Scheduler

Executor

Task Launch

ExecutorExecutor

Task Status

Scheduler

Executor

Task Launch

ExecutorExecutor

Task Status

Distributed System A Distributed System B Datacenter Operating System Approach

Distributed Systems A+B+C+...

Apache Mesos Two-level Scheduling

© 2015 Mesosphere, Inc. All Rights Reserved.

Datacenter Operating System (DC/OS)

DC/OS

...Agent 1 Agent 2 Agent 3 Agent n

Server VM Cloud VM

...Modern App Components

FROM SERVER TO DATACENTER-SCALE

Server Operating System

OS (e.g., Windows, Linux)

...Core 0 Core 1 Core 2 Core n

Server

Process

© 2016 Mesosphere, Inc. All Rights Reserved. 23

MODERN APPS WITH THE DC/OS MODEL

Traditional Approach

Big Data Analytics

Stateful Service

DC/OS Model

Container App

Container App

CaaS PaaS

● Turnkey install of datacenter-wide services● Simplified operations & efficiency● Lowers barriers to experiment with new tech (e.g., Spark)

Container App

Datacenter-scale operating system

© 2016 Mesosphere, Inc. All Rights Reserved.

BRINGING MODERN APPS TO EVERYONE

● 100% Open source

● No limits

25

THE DC/OS COMMUNITY

© 2016 Mesosphere, Inc. All Rights Reserved.

DC/OS Operations

© 2016 Mesosphere, Inc. All Rights Reserved.

DC/OS SERVICE INSTALL:SPARK

© 2016 Mesosphere, Inc. All Rights Reserved.

DC/OS SERVICE INSTALL:KAFKA

© 2016 Mesosphere, Inc. All Rights Reserved. 29

DC/OS SERVICE UPDATE & CONFIG:KAFKA

© 2016 Mesosphere, Inc. All Rights Reserved.

DC/OS UNIVERSE: STORE FOR APP SERVICES

DC/OS

Anyone can publish on DC/OS Universe app store - partners & OSS contributors

Popular services include Spark, Cassandra, Jenkins, and Kafka

Services include distributed systems that run elastically across the datacenter

Install these services with a single command

Spark

Jenkins

Riak

DataStax Enterprise

Confluent Kafka

ArangoDB

GitLab

Cassandra

JFrog Artifactory

Elasticsearch

MariaDB

Storm

HDFS

Zeppelin

MemSQL

Over 40 Services Made For DC/OS

© 2016 Mesosphere, Inc. All Rights Reserved.

TRY DC/OS IN UNDER 20 MINUTES

dcos.io

© 2016 Mesosphere, Inc. All Rights Reserved.

MESOSPHERE ENTERPRISE DC/OS

© 2016 Mesosphere, Inc. All Rights Reserved.

“We’re running real time analytics and visualization of data from connected sensors

Before DC/OS we had no where near the scale we had today, from thousands of to millions of events per second.

Next, we’re doing predictive GIS”

Adam Mollenkopf, Real Time Geospatial Information Systems Capability Lead

© 2016 Mesosphere, Inc. All Rights Reserved.

“We are building a PaaS and services on an existing OpenStack cloud to enable developers to build microservices on Docker containers and use big data technologies...

DC/OS was a force multiplier for us to get these new services out the door fast...

We transitioned the full enterprise infrastructure at 2X the speed of previous efforts.”

Tim PletcherEngineering Director, Application Services

© 2016 Mesosphere, Inc. All Rights Reserved.

“We came across Mesosphere while planning for the coming IoT market..

If I build everything in a silo, I have no chance...

With DC/OS, I have one single contiguous cluster and I can bring my data and store it and I can run all my applications as well..

and I think that's a huge advantage”

Larry Rau, Director Architecture & Infrastructure, Verizon Labs

© 2016 Mesosphere, Inc. All Rights Reserved.

SPEED MATTERS

McKinsey Quarterly 2015: “The perils of ignoring software development”Based on assessment of software development projects at 1,300 companies

Top quartile companies bring new software features to market

3x faster than bottom quartile

© 2016 Mesosphere, Inc. All Rights Reserved.

Thank YouQuestions?

@edwardihsu Learn more at mesosphere.com, including our whitepaper:

Modern Enterprise App Operations with DC/OSLessons from Running Containers, Microservices, and Stateful Big Data Services in Production