+ All Categories
Home > Technology > September16

September16

Date post: 09-Feb-2017
Category:
Upload: mark-burgess
View: 779 times
Download: 0 times
Share this document with a friend
28
How can we know when a system is working well? @markburgess_osl 2016
Transcript

How can we know when a system is working well?

@markburgess_osl 2016

Fault-resilience as promise theory project

An attempt to understand and formalise

Very hard to write down true statements about systems, and provide non-use-case specific advice.

My attempt to make progress:

http://markburgess.org/Faults.pdf

Five key words

• Who are “we”? — it’s subjective

• What does “know” mean? — know it like a friend

• What is “the system”? — where does it start and end?

• What is “working”? — what’s the basis for judgement?

• What is “well”? — can we quantify it?

“How can we know when a system is working well?”

“System” - the modularity misconception

• In a “system”, there are freedoms and constraints in balance.

• Is it one thing, or many components working together?

• NB,this is a question of scale and perspective.

• Microservices?

• Causation becomes entangled - the modularity myth

Things to understand in systems

• What is intended / what is claimed (promised) - fit for purpose

• What is actual / what is measured (assessed) - are we succeeding?

• My goals might be different from yours (subjective)

• Only the origin agent is authoritative about its intent (autonomy)

• How do assessments change with scale and perspective (relativity)

Understanding “What” happens (2000-2003)

• On what scale?

• Space • Time

• Assessments … uncertainty

• Non-deterministic

• Dynamics, but something is missing …

“Well” - dynamically and semantically suitable

• It meets “our” expectations • Who is “us”?

• Expectations are based on • Assumptions • Whose assumptions?

• The origin agent is the authoritative source • But the receiver is responsible

DESIGNING SYSTEMS FOR COOPERATION

MARK BURGESS

THINKING INPROMISES

“This is where a great quote goes. Excellent book!” –Joe Blough

“We” - the subjective (2004-2017)

• Fitness for purpose

• Desired outcome

• Forms the measuring stick for correct outcome

• Define agent

• Define promise

• Define assessment

The basics of promise theory • An agent can only make a promise about its own behaviours

• Obliging others is an ineffective strategy.

• An agent can only assess others’ promises from its own perspective

• Any reliance (dependency) on another agent invalidates a promise

“Working” - faults and resilience (2015-)

• To define a fault, error, flaw, you have to know what was intended

• But at what scale?

• What measuring stick? • Agents and promises • What are the consequences of

promises not kept?

Agent fidelity: practical definitions

What helps/prevents agents from keeping promises?

Can a system keep all its promises and still be unpredictable?

“Know” system = continuous assessment (1999-2000)

Relationships are characterized by distributions (1999)

What is the promise here?

Cognitive computing - realtime machine learning

Knowledge management and scaling - (2007-2010)

The knowledge ladder transforms: time —> space

experiment - customize - productisation - commoditisation - utilities

Can we still still keep our promises if we scale the system?

• Change of spacetime scale • Space/bigger • Time/faster

• Scaling is more than “make it bigger” • Dynamics and semantics • Do we trade functionality for size?

How do we know a system is working well?

• Know - a relationship between the system curator and its agents

• System - a collection of agents collaborating by promises

• Working - Promises made and kept

• Well - what stats about promise keeping?

If we can’t formalise these, we can’t answer this question

@markburgess_osl

DESIGNING SYSTEMS FOR COOPERATION

MARK BURGESS

THINKING INPROMISES

“This is where a great quote goes. Excellent book!” –Joe Blough

http://markburgess.org/Faults.pdf

Extras

Does it scale? — single agents and queues

• Wind-tunnel (dynamics)

• Dimensionless ratios - universal scaling

• 1 dimensional bias: queues

Scaling of agency - semantic spacetime (2014)

Modularity is not a scalable concept

Scaling to optimize…what?

• Microservices / modularity - optimizes human knowledge

• Monolith - optimizes localization

• Continuous delivery - optimises convergence

• What costs the most?

• What optimizes certainty?

Repair versus redundancy Continuous delivery vs fault avoidance

• Rapid repair cycle is the best strategy for temporal continuity

• Repair does not change the bulk scaling assumptions - using time agility instead of bulk for resilience (space —> time)

• Local time trumps space, because space is non-local time (!)


Recommended