the new distributed application infrastructure

The New Distributed Application Infrastructure

Gordon Haff @ghaffTechnology EvangelistRed Hat

Santa Clara, CA USADecember 2015 1

This is all very new


2007

2011

2013

2014

Computing as puntuated equilibria

Rapidly changing environment

Open source innovation and recombinations

Intersecting trends Hard to predict


Traditional infrastructure & apps


Server, storage, & networking hardware

Operating system

Application& dependencies



Vertical scaling



Operating system

Application& dependencies Application

& dependenciesApplication

& dependencies

(One of) the problems



Operating system


Enter hardware virtualization



Hypervisor (or other partitioning methods)




Operating system

Operating system

Operating system

Tweaking hardware virtualization


Server, storage, & networking hardwarewith CPU, memory, I/O virtualization assists

Hypervisor (or other partitioning methods)




Operating system

Operating system

Operating system

What’s changed really?

Application components still installed within OS

Applications still long-lived & stateful

Applications still “pets”


Which was sort of the point

Server consolidation reduced CAPEX

Without (at least initially) much impact on operational model


2006

Well, that worked


The discontinuity hits

“Software is eating the world” Mobile DevOps Reimagined componentized

architectures Scale


Adopting an “ant” model for apps

Stateless (often) “Small” components Single function Expose an API Replacable cogs Portable across hybrid

infrastructures


What does an infrastructure for ants look like?


From servers to resource pools

Software-defined “everything”

Dynamic resource pool

May be provided by public cloud


Virtualized pool of IaaS resources

“Commodity” server



Operating system

Operating system

Operating system

Operating system

Software-defined storageSoftware-defined networking/NFV

OpenStack: Software-defined infrastructure


Linux

Containers: Isolation within OS

OS-level virtualization Originally BSD jails Then Solaris zones OS kernel manages

isolation, resource use, and security

Namespaces, SELinux, Cgroups in Linux


Operating system instance



Resource pool

Making containers useful

Standard packaging format

Ecosystem App composition specification

Optimized operating system foundation


Lightweight, immutable OSfor running containers



Resource pool

Container packaging forimage-based deployment

Key areas of container consistency

Container host• Isolate and secure apps on the host operating system

Container image• Package apps to include information on what's needed to run

in a container with digital signatures and encryption for security

Registry and discovery• Find and consume trusted application container images from

federated sourcesSanta Clara, CA USADecember 2015 19

Managing at scale as a single entity


Lightweight, immutable OSfor running containers



Resource pool

Container packaging/API forimage-based deployment

Resource management

Orchestration

Everyone is scaling

Not just unicorns and mammoths Three main use cases:

• Large scale workloads • Diverse workloads • Complex resource management

Grid computing reimagined?


What does scale mean?

How many systems? A big cluster or many small ones? Throughput? Scheduling frequency? How much availability required?


Handling diverse workloads

Conventional or cloud native? What type of workloads?

• CI/CD (e.g. Jenkins) • Data analytics (e.g. Spark, Storm) • Batch (e.g. Chronos) • Workflows

Single or multiple clusters?


What policies are needed?

Compliance Multi-tenancy Dependency management Avoiding repeated failures Persistent volume services Dynamic reservations


Needs for integrating with existing IT

96% see open source as an enabler of cloud native integration and conventional app modernization.

Source: Red Hat Modernization Strategies SurveyIDC September 2015


Structured & unstructured data integrationBusiness process automationModel-driven process managementEnterprise service bus & APIs

Open questions

Role of hardware virtualization On-premise vs. public cloud trends Microservices Resource management/orchestration layer(s) The post-NIST service model Favored approaches for consumption


How evenly distributed will the future be?

The future is already here—it's just not very evenly distributed.

William Gibson


Credits


Fractal: https://www.flickr.com/photos/fractal_ken/3996156539/ Flickr Creative Commons license

Punctuated equilibrium: University of California at Berkeley

Mainframe: "IBM 704 mainframe" by Lawrence Livermore National Laboratory. Licensed under Attribution via Commons - https://commons.wikimedia.org/wiki/File:IBM_704_mainframe.gif#/media/File:IBM_704_mainframe.gif

Ants: https://www.flickr.com/photos/pondapple/6502194585 Flickr Creative Commons license

Meteor, galaxy: NASA

Datacenter: Google

Dogs: https://www.flickr.com/photos/ulster/3250246355 Flickr Creative Commons license

https://www.flickr.com/photos/fractal_ken/3996156539/

https://commons.wikimedia.org/wiki/File:IBM_704_mainframe.gif

https://www.flickr.com/photos/pondapple/6502194585

https://www.flickr.com/photos/ulster/3250246355