Sie sind auf Seite 1von 37

Introduction

The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

A Real-Time Scheduler to Enhance Cache


Performance(ENCAP) in Multicore Platforms
P.Pavan Kumar (12ph0515)

13th Research Review Meeting (RRM)


Under the esteemed Guidence of
Dr.Ch. Satyanarayana
(Supervisor)
Professor and DAP
Dept. of C.S.E
JNTU-Kakinada

Dr. A. Ananda Rao


(Co-Supervisor)
Professor and DAP
Dept. of C.S.E
JNTU-Anantapur

Department of Computer Science and Engineering,


JNTU-Anantapur,
January 18, 2017
P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Outline
1

Introduction
Multicore Architecture

The idea
Proposed approach

Literature review
Related Work

Methodology
ENCAP
ENCAP Heuristics
Assumptions

Evaluation
SESC
CACTI
LITMUS(RT)

6
7

Result Analysis

8
9

Future Work

Case Study
Multi-core Distributed Cluster
Experimental Evaluation
References
P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Multicore Architecture

Multicore Architecture
Contain two or more processors on a single integrated circuit (IC)
Adopted by most chip manufacturers
Such chips contains On-Chip Caches
Shared by some (or) all of the cores on the chip
Scheduling methods should aware of these caches
In Multicore architecture all cores are
X Symmetric
X Single threaded
X Shared an L2 cache
This paper abode the issue of enhancing the shared cache memory
in the context of soft real-time system
P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Multi-core Architecture ...

Multicore Architecture

Contd.

Figure: Multi-core Architecture

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Multi-core Architecture ...

Multicore Architecture

Contd.

Figure: Multi-core Architecture

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Multicore Architecture

Applications
Database servers
Web servers (Web commerce)
Compilers
Multimedia applications
Scientific applications(CAD/CAM)
MPEG-2
High performance computing(GPU)

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Proposed approach

The idea

My Idea is of two fold,


1

First, reassembling of discouraged loads(tasks) based on the


repeatedly scaling the simultaneous process.

Second, reweighting the loads and processing.

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Proposed approach

Proposed approach

Load set can encouraged to re-schedule

Which are discouraged to be co-scheduled (earlier)


Reassemble the discouraged load set (which are eligible to reschedule)
The discouraged suitable tasks are re-partitioned into fixed size
shared cache memory sets.
Reschedule the load sets based on prediction and history of
cache interference.

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Related Work

Literature review
Resources
Int. Journals(IEEE, Elsevier, ..)
Int. Conferences
Int. Symposium
Nat. Journals (CSI, ..)
Nat.Conferences
Tech. Reports(M.Tech, MS, ..)
Magazines(IEEE,IBM,..)

2004
3
0
0
0
0
0
0

2005
0
1
1
0
0
0
1

2006
3
2
1
0
0
1
0

2007
1
5
2
0
0
0
0

2008
1
2
3
0
0
0
0

2009
3
2
0
0
0
0
0

2010
2
1
1
0
0
0
0

2011
1
3
1
0
0
0
0

2012
2
1
2
0
0
0
0

2013
0
2
2
0
0
0
0

2014
0
1(S)
1
0
0
1
0

Table: Literate review(2004-2015)


*** S= In Submission
+
+
+
+
+

Total No of Papers Referred: 53


No of National/International Journals Referred: 16
No of National/International Conferences Referred: 20
No of National/International Symposiums Referred: 14
No of other papers Referred: 03
P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Related Work

Related Work
Sr.No. Year
1
2000

Author(s)
J. Anderson et al.

Methodology
p-fair Scheduling

2
3

2002
2004

Rohot jain et al.


A.Fedorova et al.

2006

A.Fedorova et al.

2006

John M. Calandrino
et al.

Co.Scheduling
Fair grained multi
threadeing
Cache -fair algorithms
Mega Tasks

2006

A Srinivasan et al.

2007

John M. Calandrino
et al.

Optimal-Rate
based Scheduling
Periodic Real-time
task Scheduling

Objective
Effective utilization of
resources
Performance
Performance Enhancement
Unique cache allocation
Effective utilization of
resources
Effective utilization of
resources
Provide good performance for non-realtime
tasks

Limitations
Not for multicore
Non-multicore
Not for large scale
systems
Specific to some
Architectures
Need better methods for grouping of
tasks
not for cache improving
For periodic tasks
only

Table: Related Work(2000-2015)


P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Related Work ...

Related Work

Contd.

Sr.No. Year
8
2007

Author(s)
John M. Calandrino
et al.

Methodology
Global Schedulong

2008

Heuristics

10

2009

11

2010

John M. Calandrino
et al.
John M. Calandrino
et al.
Sergey Zhuravlev et
al.

12

2012

Pengcheng Nie et al.

ESHMP Scheduling

Cache-Aware RealTime Scheduling


Thread Scheduling

Objective
Achiving better performance for large scale
systems
To improve Cache performance
To Enhance the L2Cache Performance
Contention for shared
resources on multicore
processors
To achive accuracy and
scalability

Limitations
restricted to one
type of problems
Oveheads should
be minimized
Cache reuse can be
better estimated
Affect
performance
degradation positively(or)
negatively
For Heterogeneous
multicore Systems
only

Table: Related Work(2000-2015)

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Related Work ...

Related Work

Contd.

Sr.No. Year
13
2013

Author(s)
Bryan C. Ward et al.

14

2013

Jeremy P. Erickson et
al.

15

2013

Bryan C. Ward et al.

Methodology
Cache locking and
Cache scheduling
Global Schedulong
by spliting jobs

Objective
Making shared cache
more predictable
Reducing tardy ness

Global
-EDFLike(GEL)
Scheduling

Setting priority points in


soft real-time systems

Limitations
Theoretical
scheduling
Examin how to account for suspention based locking
protocol
survey paper

Table: Related Work(2000-2015)

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

ENCAP
ENCAP Heuristics
Assumptions

ENCAP (ENhancing CAche Performance)

ENCAP approach is a generalized one which reschedule the loads


which are neglected tasks from co-schedule.
Subsequent-sporadic loads on M processors are considered to be
effective at the span s

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

ENCAP
ENCAP Heuristics
Assumptions

ENCAP (ENhancing CAche Performance) Heuristics


A load set should have been discouraged from co-scheduling and the
size of the load set should not be less than the fixed storage region
set (SRS) of re-partitioned load.
More than one re-partitioned discouraged loads are not allowed to
execute its job within a core.
Time constraint of each sub load should be the frame bounded with
SS-frame span s.
Tardiness of each discouraged loads should be bounded.
All the partitioned discouraged load set will have a fixed storage
region set (SRS) of parallel threaded tasks (PTTs).

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

ENCAP
ENCAP Heuristics
Assumptions

Assumptions
we presume that load set is a set of many threaded tasks Ti ,
Where i=1,2,...n.
We assume that tasks are Organized Parallel Threaded
Tasks(PTTs).
Each PTT has a Stable Storage Region Set (SRS) that is shared by
all tasks within it.
The SRS is the region of memory referenced by tasks within a PTT the
size of this portion is called as the task size (TS) of the PTT.
PTTs are useful for specifying groups of mutual tasks.
P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

ENCAP
ENCAP Heuristics
Assumptions

Re Co-Scheduling

(a)

Original priority Points

(b)

Modified priority Points

Figure: Modifying Priorities

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Re Co-Scheduling ...

(a)

ENCAP
ENCAP Heuristics
Assumptions

Contd.

Original priority Points

(b)

Re Co-scheduling

Figure: Re Co-Scheduling

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

SESC
CACTI
LITMUS(RT)

Implementation IN SESC

Figure:

Implementation in SESC

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Implementation in SESC ...

Figure:

SESC
CACTI
LITMUS(RT)

Contd.

allocation of Jobs to SESC

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

SESC
CACTI
LITMUS(RT)

Implementation in CACTI

Figure:

Configuration file submission to CACTI

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Implementation in CACTI ...

Figure:

SESC
CACTI
LITMUS(RT)

Contd.

Result file of CACTI

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

SESC
CACTI
LITMUS(RT)

Implementation in LITMUSRT

Figure:

Performance Enhancements (Access Time and Cycle time estimation) of ENCAP on LITMUSRT on

variable data sizes


P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

SESC
CACTI
LITMUS(RT)

Implementation in LITMUSRT ...

Figure:

Contd.

File selection Performance Enhancements (Access Time and Cycle time estimation) of ENCAP on

LITMUSRT on variable data sizes


P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Result Analysis

Table:

L2 cache miss ratio per load set(Max., Min., Avg.) and cache size of 14680064 bytes in ns

Sr.No.
1

Name Of Load
Set
NORMAL

TINY NORMAL

ONE COMBO

Pfair
89.35%
(1.65,1.65,1.65)
59.35%
(2.65,1.65,1.35)
11.06%
(7.55,4.65,3.29)

P.Pavan Kumar (12ph0515)

Megatask
2.18%
(10.78,10.74,10.25)
2.88%
(11.68,11.64,11.66)
0.88%
(6.8,5.9,5.9)

ENCAP
15.36%
(3.85,3.79,3.53)
17.63%
(1.85,1.81,1.82)
14.333%
(2.45,2.41,2.43)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Result Analysis...

Contd.

Figure: Access time for shared cache memory size of 14680064 bytes in ns

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Result Analysis...

Contd.

Figure: Cycle time of shared cache memory size of 14680064 bytes in ns

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Result Analysis...

Contd.

Figure: Access time and Cycle time of shared cache memory size of
14680064 bytes in ns
P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Multi-core Distributed Cluster


Experimental Evaluation

Implementation of Multi-core Distributed Cluster

Figure:

Proposed Architecture of Multi-core Distributed Cluster

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Multi-core Distributed Cluster


Experimental Evaluation

Implementation of Multi-core Distributed Cluster

Figure:

Performance Enhancements (Access Time and Cycle time estimation) of Multi-core Distributed

Cluster on LITMUSRT on variable data sizes


P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Multi-core Distributed Cluster


Experimental Evaluation

Implementation of Multi-core Distributed Cluster

Figure:

Performance Enhancements (Access Time and Cycle time estimation) of Multi-core Distributed

Cluster on LITMUSRT on variable data sizes


P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Multi-core Distributed Cluster


Experimental Evaluation

Experimental Evaluation of Multi-core Distributed Cluster


Cycle time for shared memory in ns

Access time for shared memory in ns

35

30

2Core
4Core
8Core
16Core
32Core

30

2Core
4Core
8Core
16Core
32Core

25

25
Access Time

Cycle Time

20
20

15

15

10
10
5

0
10

80

Cycle Time for # of Tasks and # of Cores

60

(a)

40

20

0
10

80

60

40

20

No. of Tasks

No. of Tasks

(b)

Access Time for # of Tasks and # of Cores

Figure: Cycle time and Access time plot for 300 KB Storage Region Set
(SRS)
P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Multi-core Distributed Cluster


Experimental Evaluation

Experimental Evaluation of Multi-core Distributed Cluster


Cycle time for shared memory in ns

Access time for shared memory in ns

35

30

2Core
4Core
8Core
16Core
32Core

30

2Core
4Core
8Core
16Core
32Core

25

25
Access Time

Cycle Time

20
20

15

15

10
10
5

0
10

80

Cycle Time for # of Tasks and # of Cores

60

(a)

40

20

0
10

80

60

40

20

No. of Tasks

No. of Tasks

(b)

Access Time for # of Tasks and # of Cores

Figure: Cycle time and Access time plot for 600 KB Storage Region Set
(SRS)
P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Multi-core Distributed Cluster


Experimental Evaluation

Experimental Evaluation of Multi-core Distributed Cluster


Cycle time for shared memory in ns

Access time for shared memory in ns

35

30

2Core
4Core
8Core
16Core
32Core

30

2Core
4Core
8Core
16Core
32Core

25

25
Access Time

Cycle Time

20
20

15

15

10
10
5

0
10

80

Cycle Time for # of Tasks and # of Cores

60

(a)

40

20

0
10

80

60

40

20

No. of Tasks

No. of Tasks

(b)

Access Time for # of Tasks and # of Cores

Figure: Cycle time and Access time plot for 900 KB Storage Region Set
(SRS)
P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Future Work

* Minimizing system overheads


* Desiging ENCAP Frame work

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

References
+ X. Xiang, Bin Bao, Chen Ding and Kai Shen. Cache Conscious Task Regrouping on Multicore Processors.
Proc. of the 12th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing,
pp.603-611, IEEE/ACM, 2012.
+ J.Calandrino and J. Anderson. Cache-Aware Real-Time Scheduling on Multicore Platforms: Heuristics and
a Case Study. Proc. of the 20 th Euromicro Conference on Real-Time Systems, pp.299-308, IEEE, 2008.
+ J.Calandrino and J. Anderson. On the Design and Implementation of a Cache-Aware Multicore Real-Time
Schedule Proc. of the 21 st Euromicro Conference on Real-Time Systems, pp.194-204, IEEE, 2009.
+ N. Ramasubramaniam, V.V. Srinivas, and P. Pavan Kumar. Understanding the Impact of Cache
Performance on multicore Architectures. Proc. of International conference on Advances in Information
Technology and Mobile Communication(AIM-2011), pp.403-406, Springer-LNCS, 2011.
+ A. Bastoni, B.Brandenburg, and J.Anderson. An Empirical Comparison of Global, Partitioned, and
Clustered Multiprocessor EDF Schedulers. Proc. of 31st Real-Time Systems Symposium (RTSS),
pp.14-24, IEEE, 2010.
+ John M. Calandrino,Dan P. Baumberger, Tong Li, and James H. Anderson. Soft Real-Time Scheduling
on Performance Asymmeric multocore Platforms. In Proc. of the the 20th IEEE Real-Time and
Embeeded Technologyand Applications Symposium(RTAS07), 2007.
+ U. Devi and J. Anderson. Tardiness Bounds under Global EDF Scheduling on a Multiprocessor.
Real-Time Systems , Volume 38, Issue 2, pp. 133-189, February 2008.

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

References ...

Contd.

+ A. Srinivasan and J. Anderson, Optimal Rate-based Scheduling on Multiprocessors . Journal of


Computer and System Sciences , Volume 72, Issue 6, pp. 1094-1117, 2006.
+ C. Liu and J. Anderson, Supporting Soft Real-Time Parallel Applications on Multiprocessors . Journal
of Systems Architecture, Volume 60, Issue 2, pp. 152-164, February 2014.
+ U. Devi and J. Anderson, Improved Conditions for Bounded Tardiness under EPDF Pfair Multiprocessor
Scheduling . Journal of Computer and System Sciences , Volume 75, Issue 7, pp. 388-420, November
2009.
+ John M. Calandrino and James H. Anderson On the design and Implementation of a Cache-Aware
Multicore Real-Time Scheduler. In Proc. of the 21st Euromicto Conference on Real-Time Systems
(ECRT09), pp. 194204 , 2009.
+ John M. Calandrino, James H. Anderson, and Dan P. Baumberger A Hybrid Real-Time Scheduling
Approach for Large-Scale Multicore Platforms . In Proc. of the 19th Euromicto Conference on Real-Time
Systems (ECRT07), 2007.
+ Aaron Block, Bjorn Brandenburg, James H.Anderson, and Stephen Quint An Adaptive Framework for
Multiprocessor Real-Time Syatems. In Proc. of the 20th Euromicro Conference on Real-Time Systems
(ECRT08), pp. 2333 , 2008.

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Introduction
The idea
Literature review
Methodology
Evaluation
Result Analysis
Case Study
Future Work
References

Any Questions ... ?

P.Pavan Kumar (12ph0515)

A Real-Time Scheduler to Enhance Cache Performance(ENCAP)

Das könnte Ihnen auch gefallen