from plaything to production - defrag 2015

31
Duncan Johnston-Wa., Cloudso3 Defrag Conference 2015 @duncanjw

Upload: duncan-johnston-watt

Post on 12-Feb-2017

786 views

Category:

Presentations & Public Speaking


0 download

TRANSCRIPT

Page 1: From Plaything to Production - Defrag 2015

DuncanJohnston-Wa.,Cloudso3DefragConference2015

@duncanjw

Page 2: From Plaything to Production - Defrag 2015
Page 3: From Plaything to Production - Defrag 2015

C21stAlchemy

OpenSourceAseeminglymagicalprocessoftransformaFon,creaFon,orcombinaFon… @duncanjw

Page 4: From Plaything to Production - Defrag 2015
Page 5: From Plaything to Production - Defrag 2015

PlethoraofPla2orms

@duncanjw

Page 6: From Plaything to Production - Defrag 2015

HybridisaDes:na:on

@duncanjw

Page 7: From Plaything to Production - Defrag 2015

VLSDistributedSystems

@duncanjw

Page 8: From Plaything to Production - Defrag 2015

@duncanjw

Page 9: From Plaything to Production - Defrag 2015
Page 10: From Plaything to Production - Defrag 2015
Page 11: From Plaything to Production - Defrag 2015
Page 12: From Plaything to Production - Defrag 2015
Page 13: From Plaything to Production - Defrag 2015
Page 14: From Plaything to Production - Defrag 2015
Page 15: From Plaything to Production - Defrag 2015
Page 16: From Plaything to Production - Defrag 2015

0018-9162/03/$17.00 © 2003 IEEE January 2003 41

C O V E R F E A T U R E

P u b l i s h e d b y t h e I E E E C o m p u t e r S o c i e t y

The Vision of AutonomicComputing

I n mid-October 2001, IBM released a manifestoobserving that the main obstacle to furtherprogress in the IT industry is a looming soft-ware complexity crisis.1 The company citedapplications and environments that weigh in

at tens of millions of lines of code and requireskilled IT professionals to install, configure, tune,and maintain.

The manifesto pointed out that the difficulty ofmanaging today’s computing systems goes wellbeyond the administration of individual softwareenvironments. The need to integrate several het-erogeneous environments into corporate-wide com-puting systems, and to extend that beyond companyboundaries into the Internet, introduces new levelsof complexity. Computing systems’ complexityappears to be approaching the limits of humancapability, yet the march toward increased inter-connectivity and integration rushes ahead unabated.

This march could turn the dream of pervasivecomputing—trillions of computing devices con-nected to the Internet—into a nightmare. Pro-gramming language innovations have extended thesize and complexity of systems that architects candesign, but relying solely on further innovations inprogramming methods will not get us through thepresent complexity crisis.

As systems become more interconnected anddiverse, architects are less able to anticipate anddesign interactions among components, leavingsuch issues to be dealt with at runtime. Soon sys-tems will become too massive and complex for eventhe most skilled system integrators to install, con-

figure, optimize, maintain, and merge. And therewill be no way to make timely, decisive responses tothe rapid stream of changing and conflictingdemands.

AUTONOMIC OPTIONThe only option remaining is autonomic com-

puting—computing systems that can manage them-selves given high-level objectives from admini-strators. When IBM’s senior vice president ofresearch, Paul Horn, introduced this idea to theNational Academy of Engineers at HarvardUniversity in a March 2001 keynote address, hedeliberately chose a term with a biological conno-tation. The autonomic nervous system governs ourheart rate and body temperature, thus freeing ourconscious brain from the burden of dealing withthese and many other low-level, yet vital, functions.

The term autonomic computing is emblematic ofa vast and somewhat tangled hierarchy of naturalself-governing systems, many of which consist ofmyriad interacting, self-governing components thatin turn comprise large numbers of interacting,autonomous, self-governing components at the nextlevel down. The enormous range in scale, startingwith molecular machines within cells and extendingto human markets, societies, and the entire worldsocioeconomy, mirrors that of computing systems,which run from individual devices to the entireInternet. Thus, we believe it will be profitable toseek inspiration in the self-governance of social andeconomic systems as well as purely biological ones.

Clearly then, autonomic computing is a grand

Systems manage themselves according to an administrator’s goals. Newcomponents integrate as effortlessly as a new cell establishes itself in thehuman body. These ideas are not science fiction, but elements of the grandchallenge to create self-managing computing systems.

Jeffrey O.KephartDavid M.ChessIBM Thomas J. Watson ResearchCenter

Page 17: From Plaything to Production - Defrag 2015

0018-9162/03/$17.00 © 2003 IEEE January 2003 41

C O V E R F E A T U R E

P u b l i s h e d b y t h e I E E E C o m p u t e r S o c i e t y

The Vision of AutonomicComputing

I n mid-October 2001, IBM released a manifestoobserving that the main obstacle to furtherprogress in the IT industry is a looming soft-ware complexity crisis.1 The company citedapplications and environments that weigh in

at tens of millions of lines of code and requireskilled IT professionals to install, configure, tune,and maintain.

The manifesto pointed out that the difficulty ofmanaging today’s computing systems goes wellbeyond the administration of individual softwareenvironments. The need to integrate several het-erogeneous environments into corporate-wide com-puting systems, and to extend that beyond companyboundaries into the Internet, introduces new levelsof complexity. Computing systems’ complexityappears to be approaching the limits of humancapability, yet the march toward increased inter-connectivity and integration rushes ahead unabated.

This march could turn the dream of pervasivecomputing—trillions of computing devices con-nected to the Internet—into a nightmare. Pro-gramming language innovations have extended thesize and complexity of systems that architects candesign, but relying solely on further innovations inprogramming methods will not get us through thepresent complexity crisis.

As systems become more interconnected anddiverse, architects are less able to anticipate anddesign interactions among components, leavingsuch issues to be dealt with at runtime. Soon sys-tems will become too massive and complex for eventhe most skilled system integrators to install, con-

figure, optimize, maintain, and merge. And therewill be no way to make timely, decisive responses tothe rapid stream of changing and conflictingdemands.

AUTONOMIC OPTIONThe only option remaining is autonomic com-

puting—computing systems that can manage them-selves given high-level objectives from admini-strators. When IBM’s senior vice president ofresearch, Paul Horn, introduced this idea to theNational Academy of Engineers at HarvardUniversity in a March 2001 keynote address, hedeliberately chose a term with a biological conno-tation. The autonomic nervous system governs ourheart rate and body temperature, thus freeing ourconscious brain from the burden of dealing withthese and many other low-level, yet vital, functions.

The term autonomic computing is emblematic ofa vast and somewhat tangled hierarchy of naturalself-governing systems, many of which consist ofmyriad interacting, self-governing components thatin turn comprise large numbers of interacting,autonomous, self-governing components at the nextlevel down. The enormous range in scale, startingwith molecular machines within cells and extendingto human markets, societies, and the entire worldsocioeconomy, mirrors that of computing systems,which run from individual devices to the entireInternet. Thus, we believe it will be profitable toseek inspiration in the self-governance of social andeconomic systems as well as purely biological ones.

Clearly then, autonomic computing is a grand

Systems manage themselves according to an administrator’s goals. Newcomponents integrate as effortlessly as a new cell establishes itself in thehuman body. These ideas are not science fiction, but elements of the grandchallenge to create self-managing computing systems.

Jeffrey O.KephartDavid M.ChessIBM Thomas J. Watson ResearchCenter

Self-ManagingSystems

Page 18: From Plaything to Production - Defrag 2015

Selfconfiguring

@duncanjw

Page 19: From Plaything to Production - Defrag 2015

Selfop:mizing

@duncanjw

Page 20: From Plaything to Production - Defrag 2015

Selfhealing

@duncanjw

Page 21: From Plaything to Production - Defrag 2015

Selfprotec:ng

@duncanjw

Page 22: From Plaything to Production - Defrag 2015

Monitor

Analyze Plan

ExecuteState

Sensors Effectors

EnrichDelegate

AutonomicManager

Sensors Effectors

Escalate

ManagedElement

@duncanjw

Page 23: From Plaything to Production - Defrag 2015
Page 24: From Plaything to Production - Defrag 2015

@duncanjw

Page 25: From Plaything to Production - Defrag 2015
Page 26: From Plaything to Production - Defrag 2015

@duncanjw

Page 27: From Plaything to Production - Defrag 2015

@duncanjw

Page 28: From Plaything to Production - Defrag 2015

Run6me

Loca6ons

OSSBlueprintCatalog

Bridge

PAYGSubscrip:on

RESTAPI

Debugger DesignerGUI

OpenShi3BrooklynBridge

MesosBrooklynBridge

Custom

Custom

Tools

COTSBlueprintCatalog

Extensions

CustomTools

CustomExtensions

OASISCAMP OASISTOSCA

24x7Produc:onSupport

AmazonVPC

Page 29: From Plaything to Production - Defrag 2015

@duncanjw

Page 30: From Plaything to Production - Defrag 2015

Plaything>ProducFon

@duncanjw

Page 31: From Plaything to Production - Defrag 2015

@duncanjw