Cooper: Task Colocation with Cooperative Games Qiuyun Llull, - - PowerPoint PPT Presentation

▶
cooper task colocation with cooperative games
SMART_READER_LITE
LIVE PREVIEW

Cooper: Task Colocation with Cooperative Games Qiuyun Llull, - - PowerPoint PPT Presentation

Mar 21, 2024 194 likes •584 views

Cooper: Task Colocation with Cooperative Games Qiuyun Llull, Songchun Fan, Seyed Majid Zahedi, Benjamin C. Lee Presented by: Qiuyun Llull Duke University HPCA Feb 7, 2017 1 Task Colocation in Datacenters Datacenters colocate applications

slide-1
SLIDE 1

Cooper: Task Colocation with Cooperative Games

Qiuyun Llull, Songchun Fan, Seyed Majid Zahedi, Benjamin C. Lee Presented by: Qiuyun Llull Duke University HPCA – Feb 7, 2017

1

slide-2
SLIDE 2

Task Colocation in Datacenters

2

Datacenters colocate applications to increase server utilization

slide-3
SLIDE 3

3

Colocation Contention

Colocation interference can lead to performance degradation

slide-4
SLIDE 4

4

System Setting

  • Alvin, Ben, and Dan are working towards HPCA papers.
  • They share a cluster and divide processors equally.
  • Ben’s applications are memory intensive.
  • Alvin and Dan’s applications are not memory intensive.
slide-5
SLIDE 5

5

System Setting

slide-6
SLIDE 6

6

Strategic Behavior

  • Alvin, Ben, and Dan are strategic.
  • Can smaller, separate clusters improve performance?
  • Alvin and Dan share separate cluster to improve performance.
slide-7
SLIDE 7

7

Strategic Behavior

slide-8
SLIDE 8

8

Without incentives, strategic users may…

  • Bypass common management policy
  • Migrate tasks for better colocations
  • Procure private machines

Strategic action fragments cluster and harms efficiency

Strategic Behavior

slide-9
SLIDE 9

9

Pursues Performance

  • Predicts contention quickly and accurately
  • Colocates tasks for system performance
  • Colocates tasks with complementary demands

Neglects Incentives

  • Overlooks strategic behavior
  • Fails to encourage users to colocate

Prior Research

slide-10
SLIDE 10

10

Stability

  • No group of users break away to form separate system

Satisfied Preferences

  • More users colocate with preferred tasks

Fair Attribution of Costs

  • Users that contribute more to contention suffer higher losses

Incentivizing Colocation

slide-11
SLIDE 11

11

Example

Preferences

A : B > C > D B : A > C > D C : A > B > D D : C > A > B

slide-12
SLIDE 12

12

Example

Preferences

A : B > C > D B : A > C > D C : A > B > D D : C > A > B

slide-13
SLIDE 13

13

A framework that incentivizes strategic users to colocate by providing desirable system outcomes:

  • Stability
  • Satisfied Preferences
  • Fair Attribution of Costs

Cooper

slide-14
SLIDE 14

14

  • System Setting
  • Incentivizing Colocation
  • Cooper Colocation Framework
  • Evaluation

Agenda

slide-15
SLIDE 15

15

  • Strategic agents are users and tasks
  • Utility is task performance
  • Colocation preferences describe preferred co-runners
  • If u(A,B) > u(A,C), then A prefers B over C
  • Actions are -- participate or break away

Cooperative Game

slide-16
SLIDE 16

16

Colocations are stable when no group of users can improve their performance by changing colocation.

Game Equilibrium

slide-17
SLIDE 17

Cooper Framework

17

Query Interface Preference Predictor Action Recommender Users Agents Coordinator System Profiler Colocation Policies Job Dispatcher Machines

slide-18
SLIDE 18

Colocation Policies

18

Query Interface Preference Predictor Action Recommender Users Agents Coordinator System Profiler Colocation Policies Job Dispatcher Machines

slide-19
SLIDE 19

Matching people in life

19

slide-20
SLIDE 20

20

Stable Matching

Algorithm partitions tasks into two sets

  • Tasks in one set propose.
  • Tasks in other set accepts, rejects.

Task updates co-runners

  • Accept proposal if performance improves

Algorithm terminates when all tasks matched

[1] D. Gale and L. Shapley, “College admissions and the stability of marriage,” American Mathematical Monthly, 1962. [2] R.W .Irving, “ An efficient algorithm for the stable roommates problem,” Journal of Algorithms, pp. 577–595, 1985.

slide-21
SLIDE 21

21

A F E D B C D>F>E E>F>D E>D>F A>B>C A>C>B C>B>A

Stable Matching

slide-22
SLIDE 22

22

Stable Policies

Stable Marriage Random (SMR)

  • Partition tasks randomly

Stable Marriage Partition (SMP)

  • Partition tasks with domain-specific knowledge
  • Memory-intensive tasks propose

Stable Roommate (SR)

  • No partition
  • Any task proposes to any other.
slide-23
SLIDE 23

23

Baseline Policies

Greedy (GR)

  • Colocate tasks to minimize performance loss

Complementary (CO)

  • Colocate tasks with complementary resource demands
slide-24
SLIDE 24

Preference Predictor

24

Query Interface Preference Predictor Action Recommender Users Agents Coordinator System Profiler Colocation Policies Job Dispatcher Machines

slide-25
SLIDE 25

25

  • Profile colocation performance with sparse samples
  • Rate co-runners with profiles
  • Predict ratings with collaborative filtering
  • Infer ratings based on task similarity
  • Suppose A: B > C and A is similar to D
  • Then D: B > C
  • Construct preference list per task based on ratings

Preference Predictor

slide-26
SLIDE 26

Action Recommender

26

Query Interface Preference Predictor Action Recommender Users Agents Coordinator System Profiler Colocation Policies Job Dispatcher Machines

slide-27
SLIDE 27

27

  • Assess assigned matches for each task
  • Search preference list for better co-runners
  • Suppose X: A > B, and X matched to B
  • X messages A to suggest new match
  • Recommend break away
  • Suppose A also prefers X over assigned match.
  • X, A should break away

Action Recommender

slide-28
SLIDE 28

Cooper Recap

28

Query Interface Preference Predictor Action Recommender Users Agents Coordinator System Profiler Colocation Policies Job Dispatcher Machines

slide-29
SLIDE 29

29

  • System Setting
  • Incentivizing Colocation
  • Cooper Colocation Framework
  • Evaluation

Agenda

slide-30
SLIDE 30

Workloads

  • PARSEC for multithreaded benchmarks
  • Spark for task-parallel machine learning

System Measurements

  • 10 nodes, each with 2 processors and 24 cores
  • Two tasks share a processor each with half the cores

System Simulation

  • 500 nodes with varied task populations
  • Simulate colocations with system profiles

30

Experimental Methods

slide-31
SLIDE 31

Fair Attribution of Costs

31

Stable Marriage Random (SMR)

swapt. bodytr. dedup caneal svm linear streamc. decision gradient naive correlat.

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35

Throughput Penalty

Complementary (CO)

s w a p t . b

  • d

y t r . d e d u p c a n e a l s v m l i n e a r s t r e a m c . d e c i s i

  • n

g r a d i e n t n a i v e c

  • r

r e l a t .

0.00 0.05 0.10 0.15 0.20 0.25

Throughput Penalty

Tasks that contribute more to contention suffer higher penalties x-axis sorts applications by memory intensity

slide-32
SLIDE 32

32

Satisfied Preferences

More users colocate with preferred tasks.

SR/CO SMR/CO SMP/CO 200 600 1000 Number of Agents Improved Performance Unchanged Performance Degraded Performance

slide-33
SLIDE 33

33

Stability

Fewer users break away to form separate system

1 2 3 4 5 200 400 600 800 CO 1 2 3 4 5 200 400 600 800 GR 1 2 3 4 5 200 400 600 800 SMR

# users breaking away x-axis: gains (%) for which tasks break away

slide-34
SLIDE 34

Loss Relative to Stand-Alone

stable matches

0.1 0.0 0.1 0.2 0.3 0.4 0.5

34

Performance

Stable colocations preserve system performance

slide-35
SLIDE 35

More in the paper …

Cooper Implementation

  • Profiler and preference predictor
  • Adapted matching algorithms
  • Action recommender and job dispatcher

Cooperative Game Theory

  • Shapely value for fair division
  • Extending beyond pairs

Experimental Results

  • Sensitivity to system scale and job mix
  • Comprehensive policy comparisons

35

slide-36
SLIDE 36

Conclusion

Cooperative Games for Shared Systems

  • Formalize interactions between strategic users
  • Incentivize user participation
  • Enable fair task colocation

Management Desiderata

  • Fair attribution of costs
  • Satisfied preferences
  • Stability

Fairness versus Performance

  • Stable colocations satisfy more users
  • Stable colocations preserve system performance

36

slide-37
SLIDE 37

Q & A

Thank you!

35