Running head: PERFOMANCE COMPARISON

Algorithms for Static Scheduling of DAGs to Multiprocessors: Performance Comparison

Student Name

Course Code

Due Date

Institutional Affiliation
PERFOMANCE COMPARISON 2

Algorithms for Static Scheduling of DAGs to Multiprocessors: Performance Comparison

Executive Summary

Difficulty mostly experienced in relation to advanced computing is founded on

scheduling computation workloads that occur primarily on heterogeneous processors. Computing

with multiprocessors holds several processing units (CPUs) that run parallel programs. Through

this, fast results are obtained from the computer, increasing its ability to manage high volumes of

data while solving problems faster.

Introduction

Heterogeneous systems exist as systems that hold a variety of resources. Difficulty

mostly experienced in relation to advanced computing is founded on scheduling computation

workloads that occur primarily on heterogeneous processors. A detailed evaluation of the

performance expected from the various algorithms will be the project's final output.

Additionally, the performance was evaluated in relation to their effectiveness when reducing

span, a process that highlights process execution and speed. The first week of the project

commenced with the determination of appropriate projects that chose a specific title. Scheduling

becomes essential because its main task focuses on breaking down an application into smaller

tasks assigned to appropriate processors to decrease the total execution time.

Difficulty mostly experienced in relation to advanced computing is founded on

scheduling computation workloads that occur primarily on heterogeneous processors. Static

scheduling focuses on details that enhance the execution of tasks, increasing communication

timings ahead of time. Through this, several basic classes of methods are employed for tacking

static scheduling in task-clustering, task-duplication, and list-scheduling.

PDG (Program dependence graph) is a frequently used cyclic graph that is employed to

mimic a sequential program existing as partitioned (DAG) (Taher et al., 2020). As such, vertex

represents specific jobs, and their weights represent the time it takes to complete the process. As

such, every edge is defined on the relationship within precedence of two positions. These

models, modes, and weights are gained through the estimation, with variations in the DAG

model described under the accurate model. Computing with multiprocessors holds several

processing units (CPUs) that run parallel programs. Through this, fast results are obtained from

the computer, increasing its ability to manage high volumes of data while solving problems

faster. These approximate models apply to the grain-size of the process, usually exceeding the

receiving and sending time. A comparison of performance between two alternative methods was

undertaken to undertake this project effectively. IPEFT and Random HEFT were compared in

performance with HEFT, a widely used technique.

Description of Hardware and Software

The DAG model under a parallel program is set under several dependencies. Through

this, indivisible units of execution are expressed under atomic models. Difficulty mostly

experienced in relation to advanced computing is founded on scheduling computation workloads

that occur primarily on heterogeneous processors. These models, modes, and weights are gained

through the estimation, with variations in the DAG model described under the accurate model.

Computing with multiprocessors holds several processing units (CPUs) that run parallel

programs. Through this, fast results are obtained from the computer, increasing its ability to

manage high volumes of data while solving problems faster. These approximate models apply to

the grain-size of the process, usually exceeding the receiving and sending time. The
PERFOMANCE COMPARISON 4

communication, in this instance, is founded on hardware to ensure that the processor reduces the

amount of time spent on communication.

Scheduling of networks occurs through direct and indirect mapping. Direct mapping of

DAG is founded on network topology through the accurate model. On the other hand, indirect

mapping is established on processor scheduling focusing on one of the models. Through this

instance, the scheduling of processors is founded on approximate models, thus ought to ignore

the network topology of the processor.

Sets of Measurements

Comparative evolution was presented in the paper within the algorithms in question and

compared to the HEFT algorithm. It is important to note that through this performance evolution

and analysis attached to HEFT, random HEFT and IPEFT algorithms were compared under

given metrics. Through the evaluation and analysis of the above algorithms, application graphs

were randomly generated, highlighting the problem. Through this, the aim of noting different

algorithmic influences on task completion is evaluated from a given dynamic acyclic graph

within a heterogeneous context. Additionally, comparison and analysis were undertaken to

provide clear terms that could reduce the time expected to translate to completion.

As an effective means to compare HEFT, Taher et al. (2020) reveal its advantages as a

significant factor. Through this, the main aim of the research will be evaluated, resulting in the

determination of resource efficiency and performance evaluation of specific real-time systems

within testing environments of scheduling methods. The first week of the project commenced

with the determination of appropriate projects that chose a specific title. Additionally, gathering

information from surrounding as stated within the requirements section was undertaken. The
PERFOMANCE COMPARISON 5

outcome, as expected, will be in line with the visual representation of the algorithm to enhance

comprehension of the project at hand. Specifically, each method is expected to provide graphs

that depend on the criteria such as jobs in question, CCR, Regularity, Maximum Jump, and

Density. The improved makespan percentage was used as a performance indicator to evaluate

these different methods effectively. Therefore, statistical analysis of the graph aided in

appreciating resource consumption findings, highlighting the best approach from those

considered under evaluation.

The first week of the project commenced with the determination of appropriate projects

that chose a specific title. Additionally, gathering information from surrounding as stated within

the requirements section was undertaken. Suman and Kumar (2018) reveal that the most

appropriate setting to undertake the coding is the Windows or Mac environment. Moreover, the

Python program language was the most preferred because it holds the potential to appreciate

many built packages and modules that make it easier for code writing to be performed and

understood. Jupyter notebook was employed to import any stored pickle files, graphically

depicting any corrupted data while comparing good data within given parameters.

Role or Job Description

The researcher will undertake heterogeneous collection processors. First, the researcher

defines the linked processors, with the importance of communication highlighted and the

completion of tasks evaluated. A detailed evaluation of the performance expected from the

various algorithms will be the project's final output. Additionally, the performance was evaluated

in relation to their effectiveness when reducing span, a process that highlights process execution

and speed.To ensure that the price is minimal for all jobs, computation of the same processors is
PERFOMANCE COMPARISON 6

critical. Through this, mapping is undertaken that projects dependent subtasks that focus on

available resources within a given period, usually the shorter period. Thomal L. Adam et al.

(1974) scheduled the node from a queue to the processor that allows the earliest execution. The

first week of the project commenced with the determination of appropriate projects that chose a

specific title. Additionally, gathering information from surrounding as stated within the

requirements section was undertaken. The time complexity in this instance will be O (V2), with

limitations within the communication time not considered. Further, Kruatrichue, B. and Lewis,

T. (1987) utilized idle time by creating a partial schedule within the processors. The limitation is

that the project failed to provide a guarantee within the optimal schedule.

Project Milestone

A detailed evaluation of the performance expected from the various algorithms will be

the project's final output. Additionally, the performance was evaluated in relation to their

effectiveness when reducing span, a process that highlights process execution and speed. The

first week of the project commenced with the determination of appropriate projects that chose a

specific title. Additionally, gathering information from surrounding as stated within the

requirements section was undertaken. After submitting the research proposal, an action plan was

developed to provide weekly updates. These updates would provide information relating to the

next week and continue the tasks until the overall task is complete. It is important to note that

from the above functions, several revisions were pointed out that necessitated the overall

completion of the project, resulting into account future phases. On a given weekend, talks

founded on the code base started were commenced. The choice of algorithms was undertaken,

which resulted in the attainment of goals. Additionally, talks founded on measurements, output,

and visual representation were also launched.

Conclusion

Parallel task scheduling is a crucial factor that aids in increasing the overall performance

within a multiprocessor system. A detailed evaluation of the performance expected from the

various algorithms will be the project's final output. Additionally, the performance was evaluated

in relation to their effectiveness when reducing span, a process that highlights process execution

and speed.Therefore, Multiprocessor scheduling can be explained through the DAG model, with

the taxonomy of DAG scheduling being a vital entity. As such, four classifications can be

obtained that are founded on algorithms. It is important to note that from the above functions,

several revisions were pointed out that necessitated the overall completion of the project,

resulting into account future phases. On a given weekend, talks founded on the code base started

were commenced. The UNC scheduling, the BNP scheduling, the TDB scheduling, and APN

scheduling. Therefore task scheduling's objective will be to mask tasks parallel to

multiprocessors that result in a reduction of their execution to result in a minimal schedule

length. Comparison of these algorithms could result in the development of parameters in the

form of makespan, speed up, and overall performance.

Bibliography

Kumar, M., Kaur, L., & Singh, J. (2020). Dynamic and static energy efficient scheduling of task

graphs on multiprocessors: A Heuristic. IEEE Access, 8, 176351-176362.

Suman, C., & Kumar, G. (2018, August). Analysis of Process Scheduling Algorithm for

Multiprocessor System. In 2018 7th International Conference on Reliability, Infocom

Technologies and Optimization (Trends and Future Directions)(ICRITO) (pp. 564-569).

IEEE.

Taheri, G., Khonsari, A., Entezari-Maleki, R., & Sousa, L. (2020). A hybrid algorithm for task

scheduling on heterogeneous multiprocessor embedded systems. Applied Soft

Computing, 91, 106202.