On-Demand Video Transcoding Using Cloud Services

The architecture for on-demand video transcoding includes includes six main components, namely video splitter, task (i.e., GOP) scheduler, transcoding virtual machines (VM), elasticity manager, video merger, and caching policy.

video splitter splits a video into GOPs, which can be transcoded independently. The code for video segmentation can be downloaded here. Also, the benchmark videos can be downloaded here.

The main contributions of this project are a QoS-aware scheduling method for task (i.e., GOP) scheduler and a dynamic resource provisioning policy for elasticity manager.

How to use it in Eclipse

Step1: Download and import project

File -> Import -> Maven -> Existing Maven Projects -> browse the location of cloudsim-projects/module

Step2: Download dependent jars through Maven

Right click module -> Maven -> Update Projects
Right click module -> Run As -> Maven build ... -> Goals ("clean install") -> Run

Project structure

This project includes four module:

  * cloudsim
  * cloudsim-example
  * cloudsim-impl
  * cloudTranscodingApp

cloudsim and cloudsim-example module are from original cloudsim package, more information can be found on the CloudSim's web site.

cloudsim-impl module mainly implements broker and datacenter for scheduling, it contains class extends class from cloudsim (e.g. TranscodingBroker extends DatacenterBroker, VideoSegment extends Cloudlet...)

cloudTranscodingApp includes the main function where the whole simulation starts. The system reading video requests through here. Resource provisioning is also implemented in this module.

How to test

Define command lines.

String[] args = {"-property", "/your/directory/to/put/config.properties",    //location to store property file
                 "-input", "/your/directory/to/put/inputdata",               //location of inputdata
                 "-output", "/your/directory/to/put/output.txt",             //location for outputdata
                 "-sortalgorithm", "SDF",          //sorting algorithm
                 "-startupqueue",                  //whether includes startup queue or not
                 "-stqprediction",                 //whether include startup queue prediction or not
                 "-videonum", "500",               //video request number
                 "-vmqueue", "1",                  //vm local queue length
                 "-vmNum", "0",                    //vm number, "0" means dynamic
                 "-vmfrequency", "10000",          //vm provisioning frequency
                 "-goplength", "AVERAGE",          //gop length
                 "-upthreshold", "0.10",           //provisioning upper threadshold
                 "-lowthreshold", "0.05",          //provisioning lower threadshold
                 "-testPeriod", "1200000",         //Test period
                 "-rentingTime", "100000",         //vm renting time
                 "-seedshift", "2"};

Download

The downloaded package contains all the source code, examples, jars, and API html files.

Publications

• Xiangbo Li, Mohsen Amini Salehi, Magdi Bayoumi, Rajkumar Buyya, CVSS: A Cost-Efficient and QoS-Aware Video Streaming Using Cloud Services, in Proceedings of 16th ACM/IEEE International Conference on Cluster Cloud and Grid Computing (CCGrid ’16), Columbia, May 2016.
• Xiangbo Li, Mohsen Amini Salehi, Magdi Bayoumi, High Performance On-Demand Video Transcoding Using Cloud Services, in Proceedings of the 16th ACM/IEEE International Conference on Cluster Cloud and Grid Computing (CCGrid ’16), Columbia, May 2016 (Doctoral symposium)
• Xiangbo Li, Mohsen Amini Salehi, Magdi Bayoumi, Cloud-Based Video Streaming for Energy- and Compute-Limited Thin Clients, Presented in Stream2015 Workshop at Indiana University, Indianapolis, USA, Oct. 2015.

Licence

Source code can be found on github.

Developed by Xiangbo Li