Thread Cluster Memory Schedulingsafari/pubs/kim_micro10_talk.pdf · Overview: Thread Cluster Memory Scheduling 1. ... Round-Robin vs. Niceness-Aware 1. ... Slide 1 Author: yoonguPublished

Thread Cluster Memory Scheduling:

Exploiting Differences in Memory Access Behavior

Yoongu KimMichael PapamichaelOnur MutluMor Harchol-Balter

Motivation

• Memory is a shared resource

• Threads’ requests contend for memory

– Degradation in single thread performance

– Can even lead to starvation

• How to schedule memory requests to increase both system throughput and fairness?

2

Core Core

Core CoreMemory

1

3

5

7

9

11

13

15

17

8 8.2 8.4 8.6 8.8 9

Max

imu

m S

low

do

wn

Weighted Speedup

FRFCFS

STFM

PAR-BS

ATLAS

Previous Scheduling Algorithms are Biased

3

System throughput bias

Fairness bias

No previous memory scheduling algorithm provides both the best fairness and system throughput

Better system throughput

Bet

ter

fair

ne

ss

Take turns accessing memory

Why do Previous Algorithms Fail?

4

Fairness biased approach

thread C

thread B

thread A

less memory intensive

higherpriority

Prioritize less memory-intensive threads

Throughput biased approach

Good for throughput

starvation unfairness

thread C thread Bthread A

Does not starve

not prioritized reduced throughput

Single policy for all threads is insufficient

thread

thread

thread

thread

Insight: Achieving Best of Both Worlds

5

thread

higherpriority

thread

thread

thread

Prioritize memory-non-intensive threads

For Throughput

Unfairness caused by memory-intensive being prioritized over each other

• Shuffle threads

Memory-intensive threads have different vulnerability to interference

• Shuffle asymmetrically

For Fairness

OutlineMotivation & Insights

Overview

Algorithm

Bringing it All Together

Evaluation

Conclusion

6

Overview: Thread Cluster Memory Scheduling

1. Group threads into two clusters2. Prioritize non-intensive cluster3. Different policies for each cluster

7

thread

Threads in the system

thread

thread

thread

thread

thread

thread

Non-intensive cluster

Intensive cluster

thread

thread

thread

Memory-non-intensive

Memory-intensive

Prioritized

higherpriority

higherpriority

Throughput

Fairness


Overview

Algorithm


Evaluation

Conclusion

8

TCM Outline

9

1. Clustering

Clustering Threads

Step1 Sort threads by MPKI (misses per kiloinstruction)

10

thre

ad

thre

ad

thre

ad

thre

ad

thre

ad

thre

ad

higher MPKI

Tα < 10%

ClusterThreshold

Intensive clusterαT

Non-intensivecluster

T = Total memory bandwidth usage

Step2 Memory bandwidth usage αT divides clusters

TCM Outline

11

1. Clustering

2. Between Clusters

Prioritize non-intensive cluster

• Increases system throughput

– Non-intensive threads have greater potential for making progress

• Does not degrade fairness

– Non-intensive threads are “light”

– Rarely interfere with intensive threads

Prioritization Between Clusters

12

>priority

TCM Outline

13

1. Clustering

2. Between Clusters

3. Non-Intensive Cluster

Throughput

Prioritize threads according to MPKI

• Increases system throughput

– Least intensive thread has the greatest potential for making progress in the processor

Non-Intensive Cluster

14

thread

thread

thread

thread

higherpriority lowest MPKI

highest MPKI

TCM Outline

15

1. Clustering

2. Between Clusters


4. Intensive Cluster

Throughput

Fairness

Periodically shuffle the priority of threads

• Is treating all threads equally good enough?

• BUT: Equal turns ≠ Same slowdown

Intensive Cluster

16

Increases fairness

Most prioritizedhigherpriority

thread

thread

thread

02468

101214

random-access streamingSl

ow

do

wn

Case Study: A Tale of Two ThreadsCase Study: Two intensive threads contending

1. random-access

2. streaming

17

Prioritize random-access Prioritize streaming

random-access thread is more easily slowed down

02468

101214

random-access streaming

Slo

wd

ow

n

7xprioritized

1x

11x

prioritized1x

Which is slowed down more easily?

Why are Threads Different?

18

Bank 1 Bank 2 Bank 3 Bank 4 Memory

rows


19

random-access


rows

•All requests parallel•High bank-level parallelism

activated row

reqreq

reqreq


20

streaming


rows

req

req

reqreq

•All requests Same row•High row-buffer locality

random-access


activated row


21

random-access streaming


rows


•All requests Same row•High row-buffer locality

stuck

Vulnerable to interference

reqreq

req

req

req

req

reqreq

TCM Outline

22

1. Clustering

2. Between Clusters



Fairness

Throughput

Niceness

How to quantify difference between threads?

23

Vulnerability to interference

Bank-level parallelism

Causes interference

Row-buffer locality

+ Niceness -

NicenessHigh Low

Shuffling: Round-Robin vs. Niceness-Aware

1. Round-Robin shuffling

2. Niceness-Aware shuffling

24

What can go wrong?




25

Most prioritized

ShuffleInterval

Priority

Time

Nice thread

Least nice thread

What can go wrong?

A

B

C

D

D A B C D

A

B

D

C

B

C

A

D

C

D

B

A

D

A

C

B

GOOD: Each thread prioritized once




26

Most prioritized

ShuffleInterval

Priority

Time

Nice thread

Least nice thread

What can go wrong?

A

B

C

D

D A B C D

A

B

D

C

B

C

A

D

C

D

B

A

D

A

C

B

BAD: Nice threads receive lots of interference





27




28

Most prioritized

ShuffleInterval

Priority

Time

Nice thread

Least nice threadA

B

C

D

D C B A D

D

A

C

B

B

A

C

D

A

D

B

C

D

A

C

B





29

Most prioritized

ShuffleInterval

Priority

Time

Nice thread

Least nice threadA

B

C

D

D C B A D

D

A

C

B

B

A

C

D

A

D

B

C

D

A

C

B


GOOD: Least nice thread stays mostly deprioritized

TCM Outline

30

1. Clustering

2. Between Clusters



Fairness

Throughput


Overview

Algorithm


Evaluation

Conclusion

31

Quantum-Based Operation

32

Time

Previous quantum (~1M cycles)

During quantum:• Monitor thread behavior

1. Memory intensity2. Bank-level parallelism3. Row-buffer locality

Beginning of quantum:• Perform clustering• Compute niceness of

intensive threads

Current quantum(~1M cycles)

Shuffle interval(~1K cycles)

TCM Scheduling Algorithm

1. Highest-rank: Requests from higher ranked threads prioritized

• Non-Intensive cluster > Intensive cluster

• Non-Intensive cluster: lower intensity higher rank

• Intensive cluster: rank shuffling

2.Row-hit: Row-buffer hit requests are prioritized

3.Oldest: Older requests are prioritized

33

Implementation Costs

Required storage at memory controller (24 cores)

• No computation is on the critical path

34

Thread memory behavior Storage

MPKI ~0.2kb

Bank-level parallelism ~0.6kb

Row-buffer locality ~2.9kb

Total < 4kbits


Overview

Algorithm


Evaluation

Conclusion

35

Fairness

Throughput

Metrics & Methodology

• MetricsSystem throughput

36

i

alone

i

shared

i

IPC

IPCSpeedupWeighted

shared

i

alone

iiIPC

IPCSlowdownMaximum max

Unfairness

• Methodology– Core model

• 4 GHz processor, 128-entry instruction window

• 512 KB/core L2 cache

– Memory model: DDR2

– 96 multiprogrammed SPEC CPU2006 workloads

Previous Work

FRFCFS [Rixner et al., ISCA00]: Prioritizes row-buffer hits

– Thread-oblivious Low throughput & Low fairness

STFM [Mutlu et al., MICRO07]: Equalizes thread slowdowns

– Non-intensive threads not prioritized Low throughput

PAR-BS [Mutlu et al., ISCA08]: Prioritizes oldest batch of requests while preserving bank-level parallelism

– Non-intensive threads not always prioritized Low throughput

ATLAS [Kim et al., HPCA10]: Prioritizes threads with less memory service

– Most intensive thread starves Low fairness

37

Results: Fairness vs. Throughput

FRFCFS

STFM

PAR-BS

ATLAS

TCM

4

6

8

10

12

14

16

7.5 8 8.5 9 9.5 10

Max

imu

m S

low

do

wn

Weighted Speedup

38


Bet

ter

fair

ne

ss

5%

39%

8%

5%

TCM provides best fairness and system throughput

Averaged over 96 workloads

Results: Fairness-Throughput Tradeoff

FRFCFS

2

4

6

8

10

12

12 13 14 15 16

Max

imu

m S

low

do

wn

Weighted Speedup

39

When configuration parameter is varied…

Adjusting ClusterThreshold

TCM allows robust fairness-throughput tradeoff

STFM

PAR-BS

ATLAS

TCM


Bet

ter

fair

ne

ss

Operating System Support

• ClusterThreshold is a tunable knob

– OS can trade off between fairness and throughput

• Enforcing thread weights

– OS assigns weights to threads

– TCM enforces thread weights within each cluster

40


Overview

Algorithm


Evaluation

Conclusion

41

Fairness

Throughput

Conclusion

42

• No previous memory scheduling algorithm provides both high system throughput and fairness

– Problem: They use a single policy for all threads

• TCM groups threads into two clusters

1. Prioritize non-intensive cluster throughput

2. Shuffle priorities in intensive cluster fairness

3. Shuffling should favor nice threads fairness

• TCM provides the best system throughput and fairness

THANK YOU

43

Thread Cluster Memory Scheduling:

Exploiting Differences in Memory Access Behavior

Yoongu KimMichael PapamichaelOnur MutluMor Harchol-Balter

Thread Weight Support

• Even if heaviest weighted thread happens to be the most intensive thread…

– Not prioritized over the least intensive thread

45

Harmonic Speedup

46


Bet

ter

fair

ne

ss

Shuffling Algorithm Comparison

• Niceness-Aware shuffling

– Average of maximum slowdown is lower

– Variance of maximum slowdown is lower

47

Shuffling Algorithm

Round-Robin Niceness-Aware

E(Maximum Slowdown) 5.58 4.84

VAR(Maximum Slowdown) 1.61 0.85

Sensitivity Results

48

ShuffleInterval (cycles)

500 600 700 800

System Throughput 14.2 14.3 14.2 14.7

Maximum Slowdown 6.0 5.4 5.9 5.5

Number of Cores

4 8 16 24 32

System Throughput(compared to ATLAS)

0% 3% 2% 1% 1%

Maximum Slowdown(compared to ATLAS)

-4% -30% -29% -30% -41%

Thread Cluster Memory Schedulingsafari/pubs/kim_micro10_talk.pdf · Overview: Thread Cluster Memory Scheduling 1. ... Round-Robin vs. Niceness-Aware 1. ... Slide 1 Author: yoonguPublished

Documents