Suggested Extension to WorkflowSim for Collaborators

For people who are interested to extend WorkflowSim and contribute back to our project, here are a few options:

1. price based scheduling (minimize cloud resource cost). WorkflowSim right now has only a few time based scheduling algorithm. This wouldn't be too difficult since we just need to put the unit price of a resource (CPU/Network/Storage) into it and run similar algorithms.

2. dynamic resource provisioning. Right now WorkflowSim first provisions VMs and delete them after the whole workflow is done. This paper is related: http://isi.edu/~gideon/publications/Malawski-Ensemble.pdf

3. multi-cluster support. we have only tested single cluster. Hybrid cloud scheduling. This paper is related: http://www.cloudbus.org/papers/Cost-EffectiveProvisionCloud-WISE2012.pdf

4. log/trace analysis, WorkflowSim requires users to specify the data size and cpu requirement before we run the simulation. Right now we support Pegasus and I am thinking to extend the support to some other popular framework if you are familiar with them and have some experience with Big Data.

5. failure aware scheduling. WorkflowSim can generate task/job failures randomly but there is no test of failure aware scheduling algorithms.

6. scheduling with resource constraints (storage/memory/network etc. ) http://isi.edu/~wchen/papers/spc-final.pdf

7. self healing/stealing algorithms. right now we have a centralized scheduler and you may change it to a distributed self healing/stealing algorithm so as to balance the load without a single node. http://charm.cs.illinois.edu/newPapers/12-11/main.pdf

8. duplication based algorithm to achieve better runtime. Right now we submit one job to one node. Maybe we can submit the same job to multiple nodes.

Feel free to do whatever you like. It is open source! If you need help, just let us know. ([email protected])