Simulation of Kanban-based scheduling for

Systems of Systems

By Alexey Tregubov, Jo Ann Lane

University of Southern California

Los Angeles, CA

Outline • Part I. Kanban-based scheduling in SoS

• SoS environments

• KSS principles

• Part II. Simulation model

• Organizational model

• Governance model

• Work items networks model

• Inputs and outputs

• Part III. Experiments and simulation results

Part I

System of Systems • Acknowledged and directed SoS:

• Three main organizational levels:

• Executive/Stakeholder management

• System Engineering team

• Product/Domain teams

Directed Acknowledged

System “a”

System “n”

System “b”

SoSE Team

System “a”

System “n”

System “b”

SoSE Team

Capabilities to products flow

C1 C2 C3 C4

R1 R2 R3 R4 R6R5

Network Team

Patient Management Team

Database Team

Cap

abili

tie

s R

eq

uir

em

en

ts

Wo

rk it

em

s

of

Pro

du

ct/D

om

ain

T

eam

s

T3 T4 T12 T2 T5

T7T9

T11T13 T1 T6 T8T10

Agg

rega

tio

n n

od

es

De

com

po

siti

on

leve

ls

C3: Telemetry Integration

C4: Performance Upgrade

R4: Performance upgrade in related systems

...R5: Integration with Patient Management Sys.

R6: Data migration

...

Database Team

T8: DB Bug-fix...

Patient Management Team

...T7: Integration Module

...

Overview In details

Health care SoS example

System of Systems’ observed issues

• Ineffective communication between different organizational levels

• Lack of visibility (status of SoS capabilities)

• Inefficient use of engineering resources

• Time wasted on context switching (multitasking overhead)

• Valuable capabilities are not delivered first or incomplete capabilities delivered:

• value delivery cadence is not satisfactory

• stakeholders cannot effectively update priorities when values change

Key measures for Kanban research

• Value delivered over time

• Schedule and effort

• Efficiency measures:

• Number of suspended/interrupted tasks

• Number of disruptive tasks

• E-factor = Uninterrupted hours / Hours present

Kanban-based scheduling process

• Eliminate waste • Minimize context switching

• Limit work in progress

• Make process more visible, transparent, and quantifiable • Kanban boards

• WIP status and obstacles

• Value-based scheduling with respect to SoS capabilities

• Efficient/explicit value control • Value-based work prioritization

• Stakeholders explicitly define value of SoS’ capabilities

• Value assigned and distributed explicitly

Executive/Stakeholder Management

Strategic planningCapability prioritization

Kanban board

TO DO:©C3: Telemetry Integration ©C4: Perfms. upgrades

WIP:©C1: Database monitoring S/W©C2: Network diagnosing S/W

DONE:©C0:Diagnostics S/W

System EngineeringAnalyze needs and capabilities.Refine Capabilities Develop requirements Allocate requirements

Form cross organization / cross disciplinary teams.Cross product and specialty engineering

Kanban board

TO DO:©R4: ...©R5: …

WIP:©R2: ...©R3: ...©R6: ...

DONE:©R1: ...

Patient Management SystemProduct SE

Identify SW featuresDecompose requirements into features

Allocate features to SW development teamsFeatures integration

Product/Domain Engineering

Pro

du

ct team

User SupportCustomer relations

Initial triage

Kanban board

TO DO:©B6: Wrong data in DB©B7: IU bug in...

WIP:©B4:Wrong date reported©B3: ...©B2: ...

DONE:©B1: Wrong midnight time format

Do

main

team

Users

Kanban board

TO DO:©T4: ...©T5: …

WIP:©T2: ...©T3: ...©T6: ...

DONE:©T1: ...

SW Development Team

Kanban board

TO DO:©T6: ...©T7: ...

WIP:©T5: ...©T4: ...

DONE:©T1: …©T2: ...

SW task development

Do

main

team

s

Kanban board

TO DO:©T4: ...©T5: …

WIP:©T2: ...©T3: ...©T6: ...

DONE:©T1: ...

Database Domain

Network Domain Kanban board

TO DO:©T4: ...©T5: …

WIP:©T2: ...©T3: ...©T6: ...

DONE:©T1: ...

Work Flow

Part II

Simulation model • Agent based model

• Three aspects of the model:

• Organizational model – structure of product and domain teams, SE team, stakeholders, etc.

• Governance model – defines agents’ behavior:

• scheduling algorithms

• queues management

• resource multitasking

• WIs and resource outsourcing policies

• WI network model – all WI and their relationships, defines:

• Work decomposition

• Value flow

Inputs and outputs

• Inputs:

• Organizational structure

• Governance model configuration

• Event scenario – events that describe how WIs originate and evolve in the simulation model.

• Outputs:

• Value delivered over time

• Number of work items in progress

• Number of fully complete & delivered capabilities over time

• Inefficiently used effort (waste of effort):

• Effort spent on context switching between tasks / multitasking

Modeling context switching • Reimmersion time:

• Modeling the reimmersion time

• Constant time: 1 hour/ 1 timeframe

• Variable reimmersion time based on

• Task complexity

• Assignment to another resource

• Length of suspension

Part III

Experiments and results

• Experiment 1 • Scheduling algorithms: KSS, LIFO, value-neutral

• Compares • value delivered over time

• total schedule and effort

• Suspended/interrupted work

• Experiment 2 • Scheduling algorithms: KSS, LIFO, FIFO, value-neutral

• Compares • value delivered over time

• Capability completeness

• Experiment 3 • KSS scheduling

• Shows impact of multitasking and work interruptions

Experiment 1

• 10 constituent teams (20 members each)

• + SoS system engineering team

• 20 new capabilities at start

• Each capability unfolds into 30 requirements on average

• Each requirement unfolds into 9 tasks on average

• Each tasks takes 3-15 days

Results: experiment 1

0

10000

20000

30000

40000

50000

60000

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80

Value

Time

KSS

Value-neutral (random selection)

LIFO

Results: experiment 1

0

5

10

15

20

25

30

35

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80

Number of Suspended Tasks

Time

KSS

Value-neutral (randomselection)

LIFO

Results: experiment 1

61

62

63

64

65

66

67

68

69

70

71

KSS Value-neutral (random selection) LIFO

Total schedule (calendar days)

Results: experiment 1

440

460

480

500

520

540

560

580

600

KSS Value-neutral (random selection) LIFO

Total effort (person-days)

Effort required if there are no interruptions

Results: experiment 2

• 15 teams (12 members each) + system engineering team.

• 10 new capabilities at start

• 20 more capabilities added during the simulation

• Each capability unfolds into 30 requirements on average

• Each requirement unfolds into 10 tasks on average.

• Each tasks takes 3-15 days.

• There are 10 expedite tasks that cause blocked work (blocked tasks)

• Simulation time-frame: 1 hour

• Simulation length (fixed time simulation): 200 days/1600 hours.

Results: experiment 2

0

20000

40000

60000

80000

100000

120000

140000

160000

180000

200000

0 20 40 60 80 100 120 140 160 180 200

Value

Time

KSS

Value-neutral (randomselection)

LIFO

FIFO

Results: experiment 2

0

5

10

15

20

25

30

KSS Value-neutral (randomselection)

LIFO FIFO

Number of 100% complete capabilities

Results: experiment 3

• We generated five groups of experiments with 100 experiments in each group. Each groups had a different number of WIs but the same organizational structure (5+1 teams). Each team had 15 resources.

• In the first group, we had 1000 WIs, in the second we had 2000 WIs, and so on.

• Nominal effort required for each WI was in interval 1-5 person-days.

• 100 disruptive WIs were introduced in the simulation.

Impact of interruptions in scale experiment 3

0

100

200

300

400

500

600

1000 WIs 2000 WIs 3000 WIs 4000 WIs 5000 WIs

Effort spent on context switching (person-days)

Conclusion and future work

Impact of Kanban process with respect to key measures

Measure

Value More value delivered over time

Effort and schedule

• Save effort on unnecessary multitasking (in simulation it reduces effort spent on context switching by 40%) • Improve capability delivery cadence

Efficiency • Reduce unnecessary interruptions and multitasking • Focus on completing capabilities (avoiding situations when everything is 90% complete and nothing delivered) • Reduce number of suspended/interrupted tasks (in simulation it reduces number of suspended tasks by 2-3 times) • Reduces the E-factor

Future work

Next steps:

• Pilot the Kanban scheduling with several organizations

• Fine-tune the simulation using empirical data and organizations’ feedback

For additional information and piloting the KSS contact:

• Alexey Tregubov tregubov@usc.edu

Q&A

• Questions?

References

• Tregubov, A. and Lane, J.A., 2015. Simulation of Kanban-based Scheduling for Systems of Systems: Initial Results. Procedia Computer Science, 44, pp.224-233.

• DeMarco, T. and Lister, T., 2013. Peopleware: productive projects and teams. Addison-Wesley.

• Boehm, B. and Turner, R., 2003. Balancing Agility and Discipline: A Guide for the Perplexed, Portable Documents. Addison-Wesley Professional.

• Dzubak, C.M., 2008. Multitasking: The good, the bad, and the unknown. The Journal of the Association for the Tutoring Profession, 1(2), pp.1-12.

• Jett, Q.R. and George, J.M., 2003. Work interrupted: A closer look at the role of interruptions in organizational life. Academy of Management Review, 28(3), pp.494-507.

Contacts

• Alexey Tregubov:

• tregubov@usc.edu

• http://atregubov.com