Vol. 2, No. 2 | July - December 2019

Enhancing Energy Efficiency in Temperature

Controlled Dynamic Scheduling Technique for Multi

Processing System on Chip

Hamayun Khan1, Muhammad Yousaf Ali khan1, Qaiser Bashir2, Muhammad Usman

Hashmi3, irfan Ud Din3 , Shahid khan4

Abstract:

1 Department of Electrical Engineering, Gomal University, D.I.Khan, KPK, Pakistan
2 Department of Electrical Engineering, The University of Lahore, Lahore, Pakistan
3 Department of Computer Science, Superior University, Lahore, Pakistan
4 United Consulting Services, D.I.Khan, KPK, Pakistan

Microprocessors designs consist of many micro level chips that reaches to a state where

thermal upsurge occurs due to rapid processing of data and effect (reduce) their efficiency in

many different aspects. That production of heat can cause disintegration which makes the chips

disable of doing many functions they are assigned to perform. Embedded devices are designed

to combine hardware and software, software integration can insert to hardware to perform

some specific function. Multicore embedded devices are in different shapes and dimension. It

has various applications on larger scale in networking and nuclear powerhouses to small

multimedia players printers, automobiles, cameras mobile handset due to higher demand of

power the energy becomes the major concern of the multicore devices for this a thermal aware

scheduling algorithm has been proposed that consider the migration of load from higher state

to that of lower state and considers all type of tasks and forecast them according to the priority

by maintaining the previous history. The proposed technique also considers various thermal

values by consulting the previous priorities of task on multicore systems. Migration policy is

used to share load from one core to another the algorithm efficiently decreases almost 3℃

temperatures at 40% utilization and the energy utilization is 221.3 J which is 3.12 % improved

as compared to the global EDF.

Keywords: Storm, Atmi, Dynamic Power Management, Dynamic Thermal Management

1. Introduction

Now a day’s systems are getting

advanced and they are characterized as they

have “real time” necessities to operate

efficiently. In such embedded machines, the

performance of the system does not consider

those results that come at the end of complete

simulation of a task in a multi-core system,

but consider the duration during which the

required results can be achieved [1]. Such

kind of systems is generally recognized as

“Real Time Systems”. These systems are

usually linked to such applications that need

Hamayun (et al) Enhancing Energy Efficiency in Temperature Controlled Dynamic Scheduling Technique for Multi-
Processing System on Chip (pp. 47 - 54)

Sukkur IBA Journal of Emerging Technologies - SJET | Volume 2 No. 2 July – December 2019 © Sukkur IBA University

to have a smooth operation for all the

applications that required reliability and

cannot afford any erroneous results in such

real time systems all the deadlines meet at the

accurate time and no chances for the system

to miss its deadline. If a deadline is missed a

failure as response can occur. These devices

are called hard real-time systems a very

common examples on larger scale for hard

real time systems are avionics systems,

textile industries, nuclear power plants, and

also those devices that operates with wire

systems established by present advanced

automobiles [2, 3]. 'Figure 1 illustrates the

block diagram of real time system'.

While on other side systems in which the

real time limit exists, but the system has the

ability to function weather a multi-core

system missed the deadlines or either releases

late. Such type of system is recognized as

“Soft real time embedded system”. These

embedded devices are safer as compared to

hard. If the time limit for each task is missed

and yet the system is functional such systems

are known as “Soft” real time systems. The

performance of the system can be slightly

disturbed and can affect the total

performance of the system. Such systems can

be used in networking of cellular systems [4].

While there is another system in which the

missing of deadline is acceptable. In this

case, the deadline of the task takes place.

Then there is no improvement in the overall

results either such systems are known as

“firm” real time such systems that can afford

delay of few seconds like a server of network

that can afford delay. A usual Real time

system can behave as hybrid. Sometime it

behaves as “hard” real time, for which all the

deadlines can meet according to their

arrangements for the efficient and safe

functionality of the system and some other

embedded aspects like “firm” and “soft” real

time where the infrequent failure is

acceptable. There are few characteristics of

RTES are as follow [5]. In real-time

embedded systems the overall accuracy of

the system depends on both the useful

accuracy of the system and also on time

constraints. Real-time systems know about

significant information of the application

running on the system. Real time system

depends on deadlines predictability and

consistency. Fault acceptance characteristics.

Hard, soft and firm are those real time

systems that are commonly used everywhere

and they are linked with a conventional

multi-core system. These are further

categorized in few other types like interactive

real time that gives response to a user while

the application is in running mode these types

of embedded devices depend on time so the

user can expect the response of the system

well on time. In such cases the user can have

a better system’s speed and have a

continuous process and have very less

chances of missing a deadline such systems

have almost the same characteristics like soft

real time embedded systems. All through in

this research work, the word “real time

embedded system” is concerned with

properties of an embedded systems and the

real time. Those devices that are in real time

embedded system category have a larger

number of common characteristics. Real time

systems have characteristics like embedded

while embedded systems usually have

characteristics like Real time on larger level

or either on smaller extent [6].

Fig 1: Block diagram of real time system.

A. Events in Real-time Systems

In Real time systems events are

categorize in two forms.

Hamayun (et al) Enhancing Energy Efficiency in Temperature Controlled Dynamic Scheduling Technique for Multi-
Processing System on Chip (pp. 47 - 54)

Sukkur IBA Journal of Emerging Technologies - SJET | Volume 2 No. 2 July – December 2019 © Sukkur IBA University

Synchronous Events: The release time

assigned to all the tasks of an application is

same.

Asynchronous Events: The release time

assigned to all the tasks of an application is

random. Figure 2 describes how a real time

system works with synchronous and

asynchronous events.

Fig 2: A simple view of real-time systems [7]

B. Function of RTES

There are few functions of real time

embedded systems. RTS, Nuclear power

plants monitoring and scheduling, EMB,

Printer, MP3 player and cell phone.

RTSEMB Cardiac pacemaker.

Fig 3: An Input/output Structure of Real time

systems

RTES are those systems for which it is

necessity to execute the set of jobs allotted to

it firmly within the pre-set limit restrictions.

Which is guaranteeing that all time-critical

tasks are handled within time is called real

time system with hard deadlines.

These tasks trust deeply on computational

analysis and data plays an important role for

achievement of the specific task to avoid

scheduling issues all the resources that are

distributed try to meet the time constraints for

the task not to miss any deadlines [8].

2. Literature Review

Real-time systems are divided into hard

and soft. The scheduling mechanism for hard

are functional for soft real-time scheduling.

Static and dynamic are the two further types

of hard real time scheduling. The scheduling

choices are usually occupied during the

compilation time of task that depends on

previous information about task set

constraints [9]. Static scheduling occurs

when the completing time for all the tasks are

in the condition where scheduling by the

scheduler finds earlier to the start of

execution.

Tasks in which the information has some

deadlines and the time for their execution is

required are called static scheduling. The

scheduling techniques occur during offline

execution of tasks are static. In difference to

this, dynamic scheduling makes quick

decisions when the task is in running state by

finding the set of tasks currently executing.

This type of scheduling affords flexibility.

Dynamic scheduling represents such

scheduling where the execution time for the

task is not continuous and changes with time

according to scheduler instructions.

During execution the process is completed

dynamically when there is a limited priority

assign to a task. In such order task are

executing according to the plan settled during

run time and deviates by the instructions

provided from the scheduler. Executions of

tasks with or without any issue are likely

with both static and dynamic scheduling.

Real time operating system proposed various

scheduling mechanisms.

All these methods for scheduling have that

issue that they allow prompt tasks to achieve

execution by disturbing the execution of

running tasks. All of them fall in the

Hamayun (et al) Enhancing Energy Efficiency in Temperature Controlled Dynamic Scheduling Technique for Multi-
Processing System on Chip (pp. 47 - 54)

Sukkur IBA Journal of Emerging Technologies - SJET | Volume 2 No. 2 July – December 2019 © Sukkur IBA University

classification of non-preemptive scheduling

[10].

There are few Scheduling characteristics that

are necessary for scheduling of a multi core

real time system that are as follow [11, 12,

and 13].

• Guaranteeing that system timing restraints
are meeting.

• Avoid synchronized installment of
devices and mutual resources.

• Attaining a state where utilization is very
high.

• Confirms that the dispatch cost of a real-
time systems must be low

Fundamentally, the scheduling problem

includes expressing a plan for the

implementation, execution and completion of

all the tasks [14]. Due to advancement in

embedded technology periodic activities

shows more computational demand on a

larger scale in a multi-core system these

activities are categories in three types in a

task model that are Periodic, A Periodic and

Sporadic.

Figure 2-9 describes the various categories

of jobs that are used in a processor' [15].

Fig 4: Categories of Task Model

C. Periodic Task

In system monitoring periodic scheduling

plays very important role periodic tasks

usually occurs during control loops. These

kinds of activities are continuously executing

on definite rates. Periodic task repeats itself

at specific periodic time interval in a

continuous manner e.g. Control loops, Sensor

reading [16]. 'Figure 5 illustrates how

periodic tasks are scheduled for the

completion of their jobs in a specified time.

For better Scheduling of task few notations

are introduced that are as follow. For generic

task sets i represents a periodic task while Ti

represents the time period of a general

periodic tasks and øi represents the phase of

task i. Tin represents the nth occurrence to a

task “i” and Ria represents the release

discharge duration for nth instance of task i

in periodic task there are few major

parameters that are used are written below.

Reaction duration of an occurrence of task is

the time in which the instance is finished.

The time starts and calculates from the start

of an instance to the discharge of a task at

that time instance it is finished.

Rak = fak - Tak-

Significant instant of a task:

The time when larger response is created at

the release of a job.

Relation when release of a task occurs:

Among two continuous instants the major

deviation of change is known as relative

release of a task.

RRJa = max| (2i, fc-ra,fc) - (2i,fc-a-ra,fc-a).

Complete discharge jitter of a Job

The instantaneous change in the duration

between a single point instance and the

maximum instant in which the deviation of a

task occurs [43].

ARJa =max (2a, fc – ra) - min (sa, fc – ra).

There are three basic categories of scheduling

algorithms that uses periodic tasks are Rate

monotonic scheduling, earliest deadline first

Scheduling and deadline monotonic

scheduling.

Fig 5: Periodic Task Model

Hamayun (et al) Enhancing Energy Efficiency in Temperature Controlled Dynamic Scheduling Technique for Multi-
Processing System on Chip (pp. 47 - 54)

Sukkur IBA Journal of Emerging Technologies - SJET | Volume 2 No. 2 July – December 2019 © Sukkur IBA University

D. A Periodic Task

A periodic task can occur at any time interval

when the task is in running mode. There is no

specific instant for the occurrence of

aperiodic task. Aperiodic task can have very

elastic deadline and sometime aperiodic task

doesn’t have any deadlines e.g. Alarm clocks

and all emergency embedded devices that can

arrive at any time instant during emergency

like robotic cars and obstacle avoidance

system [17]. Figure 6 illustrates how

aperiodic tasks are scheduled and jobs occur

at any time instant.

3. Problem Statement

The main objective and focus of this research

work is chip (reliability) that is the most

important issues in multi core embedded

technologies. High temperature causes

multiple performances and reliability issues.

Fig 6: A Periodic Task Model [18]

Task migration is the way to avoid high

temperature and improve performance.

However accurate prediction about the

coolest core and the task which needs to

migrate to control the chip temperature is the

major design issue in task migration

techniques. Our technique considers all types

of tasks including hot and cold task to

accurately predict the temperature of the core

in running mode and also use preciously

thermal history to precisely check the chip

temperature for the selection of cold core. A

task migration technique based on previous

history by considering workload of running

core to predict the future temperature and

coolest cores. This technique improves the

reliability by avoiding thermal issues [19].

4. Experimental Technique

The major components include a task

producer where user can create random task

sets in XML file that contains different

parameter. This XML input file is used as an

input for the simulation tool. The simulation

tool for real-time multiprocessor scheduling

can simulate the input file according to the

scheduling policy and developed the power

profiles according to the set of parameters

given in XML file. These Power profiles

generated from STORM can be saved in a

text output file. The XML file contains all the

information of task sets. A temperature

model can also analyze the thermal responses

of different scheduling algorithms. A

statistical testing mechanism is used to test a

many data set. These data sets are already

stored produces thermal profile when used

with some hardware constraint in thermal

model ATMI and a power model is used for

each core comprises of static and dynamic

power.

Total power = dynamic power + static power

Dynamic power = C × F × V

Therefore, C is the switching capacitance.

F is the frequency and V is the supply

voltage

Static power = Leakage current × V

Flow chart

We have proposed technique for octa core

multiprocessing Leat processor in which the

scheduler will first check the number of task

jobs as shown in figure 7.

That is running in an application. Primarily

the value of counter is zero for all 8-cores in

a multiprocessing leat processor. The

scheduler is able to determine the constraint

and values of every individual task. So, the

counter for all the cores is initially zero. The

configuration of core will be selected on the

basis of least temperature and lower Power of

consumption among all other configurations

once the core with least temperature is

Hamayun (et al) Enhancing Energy Efficiency in Temperature Controlled Dynamic Scheduling Technique for Multi-
Processing System on Chip (pp. 47 - 54)

selected its counter in running mode is

rapidly rising. In such case when the

temperature of the core in running mode is

less than the maximum allowable

temperature then the scheduler will perform

normal process of task execution.

Fig 7: Flow Chart

5. Experimental Results

In this section we discuss our experimental

results that illustrates the temperature

variation between the curves of proposed

EDF and Global EDF. Proposed EDF

considers reliability and performance

parameter so only those cores are in running

state that is in working condition. In the

beginning exponentially, temperature on chip

rises and then arrive at a steady state level. At

13% utilization factor the global EDF has 5℃
more temperature on chip as

Fig 8: Thermal Cycling under 80Mhz for CPU1 &

CPU2 at low Utilization Factor

Fig 9: Thermal cycling under 80Mhz for CPU1 &

CPU2 at high Utilization Factor

Hamayun (et al) Enhancing Energy Efficiency in Temperature Controlled Dynamic Scheduling Technique for Multi-
Processing System on Chip (pp. 47 - 54)

Fig 10: Thermal cycling under 80Mhz for CPU1,

CPU2, CPU3 and CPU4 at low Utilization Factor

Fig 11: Thermal Cycling under 80Mhz for CPU1,

CPU2, CPU3 and CPU4 at High Utilization Factor

6. Conclusion and Future Work
In this research work functional

simulation on STORM and thermal model

and proved its practicability in the context of

MPSoC. A large amount of techniques now a

days doesn’t consider the affect of

environmental temperature which can be

investigated in future work. Majority of the

current algorithms dsefines only for

homogeneous multi-core systems which can

be extensive to heterogeneous multi-core

systems.

REFERENCES

[1] Matt Bach. How ambient temperatures affect
your pc. Temperatures-Affect-Your-PC-158/,
Written on August 15, 2012.

[2] Shekhar Borkar. Design challenges of
technology scaling. IEEE Micro, 19(4):23–
29, July 1999.

[3] Thidapat Chantem, Xiaobo Sharon Hu, and
Robert P. Dick.Temperature-aware
scheduling and assignment for hard real-time
applications on mpsocs. IEEE Trans. VLSI
Syst., 19(10):1884–1897, 2011.

[4] Jian-Jia Chen, Shengquan Wang, and Lothar
Thiele. Proactive speed scheduling for real-
time tasks under thermal constraints. In
Proceedings of the 2009 15th IEEE
Symposium on Real-Time and Embedded
Technology and Applications, RTAS ’09,
pages 141–150, Washington,DC, USA,
2009. IEEE Computer Society.

[5] H. Khan, M. U. Hashmi, Z. Khan, R. Ahmad,
and A. Saleem, “Performance Evaluation for
Secure DES-Algorithm Based Authentication
& Counter Measures for Internet Mobile Host
Protocol,” IJCSNS Int. J. Comput. Sci. Netw.
Secur. VOL.18 No.12, December 2018, vol.
18, no. 12, pp. 181–185, 2018.

[6] Farzan Fallah and Massoud Pedram. Standby
and active leakage current control and
minimization in cmos vlsi circuits. IEICE
Transactions, 88- C(4):509–519, 2005.

[7] Stephen H. Gunther, Frank Binns, Douglas M.
Carmean, and Jonathan C. Hall. Managing
the impacincreasing microprocessor power
consumption. (Q1):9, February 2001.

[8] H. Khan, M. U. Hashmi, Z. Khan, and R.
Ahmad, “Offline Earliest Deadline first
Scheduling based Technique for
Optimization of Energy using STORM in
Homogeneous Multi- core Systems,” IJCSNS
Int. J. Comput. Sci. Netw. Secur. VOL.18
No.12, December 2018, vol. 18, no. 12, pp.
125–130, 2018.

[9] Khan, H., Bashir, Q., & Hashmi, M. U.
(2018). Scheduling based energy
optimization technique in multiprocessor
embedded systems. 2018 International
Conference on Engineering and Emerging
Technologies
(ICEET).doi:10.1109/iceet1.2018.8338643

[10] Joohno Kong, Sung Woo Chung, Kevin
Skadron, “Recent Thermal Management
Techniques for Microprocessors" ACM
Computing Surveys, Vol. 44, No. 3, Article
13, pp. 13:1-13:42, June 2012

Hamayun (et al) Enhancing Energy Efficiency in Temperature Controlled Dynamic Scheduling Technique for Multi-
Processing System on Chip (pp. 47 - 54)

[11] Mehdi Kamal, A. Iranfar, A. Afzali-
Kusha, M.Pedram,"A Thermal Stress-aware
Algorithm for Power and Temperature
Management of MPSoCs" In EDAA, 2015.

[12] Q. Bashir, H. Khan, M. U. Hashmi, and S. Ali
zamin, “A Survey on Scheduling Based
Optimization Techniques in Multi-Processor
Systems,” in Proceedings of the 3rd
International Conference on Engineering &
Emerging Technologies (ICEET), Superior
University, Lahore, PK, 7-8 April, 2016.,
2016.

[13] T. S. Rosing, K. Mihic, G. De Micheli,
"Power and Reliability Management of
SoCs,"In IEEE Transaction Very Large Scale
Integrated System (VLSI), vol. 15. no.4,
pp.391-403, 2007.

[14]Pratyush Kumar,Lothar Thiele,"Thermally
Optimal Stop-GO Scheduling of Task Graphs
with Real Time Constraints,"In ASP-DAC,
IEEE Press, pp. 123-128, 2011.

[15] Alexandru Andrei, Petru Eles, Zebo Peng,
Marcus T. Schmitz, Bashir M. Al Hashimi,
"Energy optimization of multiprocessor
systems on chip by voltage selection," IEEE
Transaction on VLSI, vol 50, no.3, 2007.

[16] "1965 – "Moore's Law" Predicts the Future of
Integrated Circuits". Computer History
Museum, 2007

[17] C. J. Lasance, "Thermally Driven Reliability
Issues in Microelectronic Systems: Status-
quo and Challenges". Microelectronics
Reliability, pp. 1969–1974, December 2003.

[18] H. Khan, Q. Bashir, and M. U. Hashmi,
“Scheduling based Energy Optimization
Technique in multiprocessor Embedded
Systems,” in 2018 International Conference
on Engineering and Emerging Technologies
(ICEET). doi:10.1109/iceet1.2018.8338643,
2018.

[19] H. Khan, S. Ahmad, N. Saleem, M. U.
Hashmi, and Q. Bashir, “Scheduling Based
Dynamic Power Management Technique for
offline Optimization of Energy in Multi Core
Processors,” Int. J. Sci. Eng. Res. Vol. 9,
Issue 12, December-2018, vol. 9, no. 12, pp.
6–10, 2018.

[20] G. Petrone, G. Seagnuolo, R. Teodorescu,
'"Reliability Issues In Photovoltaic Power
Processing Systems", IEEE Transactions,
July, 2008.

http://www.computerhistory.org/semiconductor/timeline/1965-Moore.html
http://www.computerhistory.org/semiconductor/timeline/1965-Moore.html
https://en.wikipedia.org/wiki/Computer_History_Museum
https://en.wikipedia.org/wiki/Computer_History_Museum