Supply Chain Network Simulator Using Cloud Computing
Abstract
Large-scale supply chains usually consist of thousands of stock keep units (SKUs) stocked at different locations within the supply chain. The purpose of this project is to develop a prototype software program that can allow the simulation of large-scale multi-echelon, multi-item supply networks using cloud-computing resources. These simulations are essentially compute-intensive Monte-Carlo experiments requiring multiple replications. Replications will be distributed across virtual machines within a cloud architecture.
Intellectual Merit
This research will contribute to the computational sciences for the simulation of large-scale supply systems within a cloud computing architecture. The project will design simulation models of large-scale supply networks for execution within a cloud-computing environment. New insights and performance evaluation of cloud computing architectures will be obtained. The feasibility of the approach will be evaluated and cloud computing will become better understood by researchers outside of the computer science community.
Broader Impact
The proposed effort will have broader impacts because companies will be able to use the results to develop better systems and software products that rely on cloud computing for applications involving this use case. Educational materials will be developed to provide how-to knowledge for other researchers and industry collaborators. The project will contribute to the research infrastructure, computing infrastructure, and the graduate student training in the State of Arkansas, a designated EPSCOR state.
Use of FutureGrid
In order to take advantage of cloud computing resources, the following execution scenario is being considered. A Java-based web service will be available to receive the input file from the user. A simple user interface will be available to permit the uploading of the file. The web service will package the workload associated with the simulation into jobs. Each job will represent one replication of the simulation. The web service will instantiate a cluster of virtual machines. Each virtual machine will be responsible for executing one or more replications of the simulation. The web service will coordinate with the virtual machines to schedule the execution of the simulation jobs and collect the results from each replication. After all replications are completed, the web service will package the results as a spreadsheet file and notify the user through email that the results are completed.
Scale Of Use
Small sets of VMs will be needed (e.g. 5-10) for short durations (1-4 hours) in order to establish feasibility of prototype and be available for testing.
Publications
Results
Project #: FG-133
Abstract:
Large-scale supply chains usually consist of thousands of stock keep units (SKUs) stocked at different locations within the supply chain. The simulation of large-scale multi-echelon supply chain networks is very time consuming. The purpose of this project is to design a cloud computing architecture to facilitate the computational performance of large scale supply chain network simulations. A Cloud Computing Architecture For Supply Chain Network Simulation (CCAFSCNS) was designed in this project, and a prototype system was developed using the computing resources in the FutureGrid. The simulations are essentially compute-intensive Monte-Carlo experiments requiring multiple replications. Replications are distributed across virtual machines within CCAFSCNS. The results show that the cloud computing solution can significantly shorten the simulation time.
Resources used in this project (which are related to FutureGrid):
1. Virtual Machine: Grid Appliance
2. Hardware Systems: Alamo Network
3. Service Access: Nimbus Cloud Client
Completed Work:
1. Customized the Grid Appliance to be Condor Server, Condor Worker and Condor Client.
2. Designed a Cloud Computing Architecture For Supply Chain Network Simulation (CCAFSCNS).
3. Developed a prototype system that implemented the CCAFSCNS with Excel, Access, Spring Framework, supply chain network simulator, FutureGrid, the Condor System, and the Grid Appliance. The virtual machines (VMs) of the Condor Worker, which is customized based on the Grid Appliance, are started in the Alamo network. These VMs are the computing resources used in the prototype system to run simulation jobs.
4. Did a computational time study on the cloud computing solution based on FutureGird:
a. Analyzed the time components used in the cloud computing solution
b. Estimated the scheduling time for a simulation request
c. Compared the simulation time spent on traditional solution and cloud computing solution and showed that the cloud computing solution can save 70% of the simulation time.
Achievements/Publications:
1. One Master project report has been submitted to fulfill the requirement for the degree of Master of Science.
2. One conference paper has been submitted to the 2012 Winter Simulation Conference.
Broader Impacts:
A Cloud Computing application capable of evaluating the performance of multi-echelon supply networks through simulation is developed in this project. This application includes a web application that can run the simulation from the cloud and a database application that helps users develop the input data and analyze the output data. Companies will be able to use the results to develop better systems and software products that rely on cloud computing for applications involving this use case. In addition, the cloud computing architecture designed in this project can be used to develop other cloud computing solutions. Also, educational materials, such as the tutorials of building the Condor System, are developed to provide how-to knowledge for other researchers and industry collaborators.
Futuregrid is a resource provider for