state-space collapse via drift conditions

31
State-Space Collapse State-Space Collapse via Drift Conditions via Drift Conditions Atilla Eryilmaz (OSU) and R. Srikant (Illinois) 06/27/22 1

Upload: armando-patrick

Post on 31-Dec-2015

16 views

Category:

Documents


0 download

DESCRIPTION

State-Space Collapse via Drift Conditions. Atilla Eryilmaz (OSU) and R. Srikant (Illinois). Goal. Motivation. Parallel servers Jobs are buffered at a single queue When a server becomes idle, it grabs the first job from the queue to serve All servers are fully utilized whenever possible. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: State-Space Collapse via Drift Conditions

State-Space Collapse via State-Space Collapse via Drift ConditionsDrift Conditions

Atilla Eryilmaz (OSU) and R. Srikant (Illinois)

04/19/23 1

Page 2: State-Space Collapse via Drift Conditions

GoalGoal

04/19/23 2

Page 3: State-Space Collapse via Drift Conditions

MotivationMotivation

04/19/23 3

• Parallel servers• Jobs are buffered at

a single queue• When a server

becomes idle, it grabs the first job from the queue to serve

• All servers are fully utilized whenever possible

Page 4: State-Space Collapse via Drift Conditions

Multiple queuesMultiple queues• Jobs arrive and

choose to join the shortest queue upon arrival

• Total queue length is the same as in the case of a single queue if jobs “defect” to a different queue whenever one becomes empty

04/19/23 4

Page 5: State-Space Collapse via Drift Conditions

Multi-Path RoutingMulti-Path Routing• Choice of paths from source to destination: route

each packet on currently least-congested path• JSQ is an abstraction of such routing scheme. It is

not possible for packets to defect from one path to another.

• Is JSQ still optimal in the sense of minimizing queue lengths?

04/19/23 5

Page 6: State-Space Collapse via Drift Conditions

Heavy-Traffic RegimeHeavy-Traffic Regime• Consider the traffic regime where the arrival rate

approaches the system capacity

04/19/23 6

Page 7: State-Space Collapse via Drift Conditions

Foschini and Gans (1978)Foschini and Gans (1978)

04/19/23 7

Page 8: State-Space Collapse via Drift Conditions

Steady-State Result for JSQSteady-State Result for JSQ

04/19/23 8

Page 9: State-Space Collapse via Drift Conditions

Lower-Bounding QueueLower-Bounding Queue

04/19/23 9

Page 10: State-Space Collapse via Drift Conditions

The Lower BoundThe Lower Bound

04/19/23 10

Page 11: State-Space Collapse via Drift Conditions

State-Space CollapseState-Space Collapse

04/19/23 11

(1,1)

q

q

Page 12: State-Space Collapse via Drift Conditions

A Useful Property of JSQA Useful Property of JSQ

04/19/23 12

Page 13: State-Space Collapse via Drift Conditions

Drift Conditions and MomentsDrift Conditions and Moments

04/19/23 13

Page 14: State-Space Collapse via Drift Conditions

Moments & State-Space Moments & State-Space

CollapseCollapse

04/19/23 14

Page 15: State-Space Collapse via Drift Conditions

The Upper BoundThe Upper Bound

04/19/23 15

Page 16: State-Space Collapse via Drift Conditions

Using State-Space CollapseUsing State-Space Collapse

04/19/23 16

Page 17: State-Space Collapse via Drift Conditions

Handling Cross TermsHandling Cross Terms

Page 18: State-Space Collapse via Drift Conditions

TheoremTheorem

04/19/23 19

Page 19: State-Space Collapse via Drift Conditions

Three-Step ProcedureThree-Step Procedure

04/19/23 20

Page 20: State-Space Collapse via Drift Conditions

Wireless NetworksWireless Networks

04/19/23 21

Page 21: State-Space Collapse via Drift Conditions

ExampleExample• Two links, four feasible rates: (0,2), (1,2), (3,1),

(3,0)

04/19/23 22

(0,2)(1,2)

(3,1)

(3,0)

Capacity Region:Set of average service rates

Page 22: State-Space Collapse via Drift Conditions

MaxWeight (MW) AlgorithmMaxWeight (MW) Algorithm

04/19/23 23

(0,2)(1,2)

(3,1)

(3,0)

Capacity Region:Set of average service rates

Arrival rates can be anywhere in the capacity region; MW stabilizes queues

Page 23: State-Space Collapse via Drift Conditions

Lower BoundLower Bound

04/19/23 24

(0,2)(1,2)

(3,1)

(3,0)

Capacity Region:Set of average service rates

Arrival rates can be anywhere in the capacity region; MW stabilizes queues

Page 24: State-Space Collapse via Drift Conditions

Heavy-Traffic RegimeHeavy-Traffic Regime

04/19/23 25

(0,2)(1,2)

(3,1)

(3,0)

Capacity Region:Set of average service rates

Arrival rates can be anywhere in the capacity region; MW stabilizes queues

.

Page 25: State-Space Collapse via Drift Conditions

State-Space CollapseState-Space Collapse

04/19/23 26

c

q

q

Page 26: State-Space Collapse via Drift Conditions

Upper BoundUpper Bound

04/19/23 27

Page 27: State-Space Collapse via Drift Conditions

TheoremTheorem

04/19/23 28

Page 28: State-Space Collapse via Drift Conditions

ImplicationsImplications

04/19/23 29

c

q

q

Page 29: State-Space Collapse via Drift Conditions

Use Beyond Heavy-Traffic Use Beyond Heavy-Traffic

RegimeRegime

• Each face of the capacity region provides an upper and lower bound

• Treat these as constraints

• From this the best upper and lower bounds can be obtainedo Similar to Bertsimas, Paschalidis and Tsitsiklis (1995),

Kumar and Kumar (1995), Shah and Wischik (2008)

04/19/23 30

Page 30: State-Space Collapse via Drift Conditions

Stability and PerformanceStability and Performance• Stability of control policies can be shown by

considering the drift of a Lyapunov function

• Setting this drift equal to zero gives bounds on queue lengths in steady-state

• But these are not tight in heavy-traffic

• The moment-based interpretation of state-space collapse and the upper bounding ideas to use this information provide tight upper bounds

04/19/23 31

Page 31: State-Space Collapse via Drift Conditions

ConclusionsConclusions• An approach to state-space collapse using

exponential bounds based on drift conditions

• A technique to use to these bounds in obtaining tight upper bounds

• Demonstrated foro JSQo MaxWeighto MaxWeight with fading is an easy extension

04/19/23 32