125 random task-graphs for multiprocessor task scheduling

Published: 15-07-2018| Version 2 | DOI: 10.17632/4fycv9td56.2
Contributor:
Hamid Reza Boveiri

Description

For a fair evaluation, a set of 125 random task-graphs for multiprocessor task scheduling are presented in this data-set. We have the following three major parameters to differentiate the shape and structure of these randomly generated task-graphs: • Size (n): The first and most contributing parameter that is the number of nodes in the given task graph. We consider five different values as {32, 64, 128, 256, and 512}. • Communication-to-Computation Ratio (CCR): Another important parameter which present us how much a graph is communication or computation intensive. All the nodes’ weights were randomly chosen from a uniform distribution with mean equal to 50 time-instance, and all the edges’ weights are randomly selected from a uniform distribution with mean equal to average-computation-cost * CCR. We consider five different values of CCR as {0.1, 0.5, 1.0, 5.0, and 10.0}, where selecting 0.1 makes the graph highly computation-intensive while 10.0 makes it highly communication-intensive. • Parallelism: Third important parameter that is the average number of children for each task in the task-graph which contributes to the connectivity of the resulted task-graph. The higher parallelism, the higher connected graph. We considered five different values of parallelism as {3, 5, 10, 15, and 20}.

Files

Steps to reproduce

There are a set of 125 random task-graphs in this dataset for the aim of multiprocessor task scheduling, and evaluating related methods. Each random task-graph is represented by an individual .txt file where the name e.g. 32N_50W_0.1CCR_3Para_65.txt is composed of 5 parts as follows: 32N: indicate that there is 32 nodes in this task-graph. 50W: all the nodes’ weights were randomly chosen from a uniform distribution with mean equal to 50 time-instance (this property is identical for all the task-graphs). 0.1CCR: Communication-to-Computation Ratio (CCR) is 0.1, where all the edges’ weights are randomly selected from a uniform distribution with mean equal to average-computation-cost * CCR (i.e. 50 * 0.1 for this task-graph). 3Para: the average number of children for each task (Parallelism) is 3. _65.txt: the last number (here 65) is an extra random number without any specific functionality, and file type is .txt. Besides, inside each task-graph file, the first data is the number of nodes in the file, followed by a list indicating the nodes' weights, and a adjacency matrix as the edges' weights. For example, if the number of nodes in the top of the file is 32, there will be a list of 32 numbers as the nodes' weights, and a 32*32 adjacency matrix as the edges' weights.