COMS E6998-6 Cloud Computing and Data Center Networking

1
COMS E6998-6 Cloud Computing and Data Center
Networking

• Sambit Sahu
• http//www.cs.columbia.edu/sambits/

2
Project Idea 1 Elasticity

• Demonstrate on-demand resource allocation
• Scenario 1
• Build and deploy an application image (say 2
  tier)
• Instrument it to monitor CPU usage
• Add another VM to the front-end when load exceeds
  a limit
• Divide the workload between two VMs
• Scenario 2
• Build various triggering capabilities that
  someone could use to implement resource scaling
• Scenario 3
• Evaluate various triggering capabilities
  available on EC2/CloudWatch for resource scaling
• Project steps
• Invoke EC2 APIs in a programmatic way from your
  Java SDK or any prog envt
• Deploy agent on VMs to get monitoring statistics
• Write a decision code to request additional VM if
  CPU load increases
• Similarly remove resource
• Expose this as a service if possible for others
  to use

3
Project 2 Monitoring

• Integrate monitoring information and provisioning
  information into one console view
• EC2 based provisioning
• Ganglia based monitoring information
• Monitoring framework
• Hypervisor level information from Cloud Watch
• VM level information from agents

4
Project 3 Management

• Change management on running VMs
• Implement changes to existing images on cloud
• You pick any management function. Add to EC2 like
  cloud or any to demonstrate this management
  functionality.
• You need to add another layer to keep relevant
  abstraction and information.
• For change management, you need to keep track of
  software stacks to verify if a change is needed,
  the application graph to understand what affects
  a change etc.

5
Project 4 Migration

• Energy conservation
• Show that by migrating VMs based on workload,
  possible to save energy consumption.
• Server consolidation
• Use migration to consolidate VMs onto a smaller
  set of machines based on time of the day workload
  pattern
• Usage based footprint control

6
Project 5 VM placement

• Where the new VM should be placed or created?
• How to place the existing VMs such that
  additional VMs can be created?
• Make this part of the provisioning logic so that
  more VMs can be supported based on sizing request
• You can take any cloud provisioning stack.
  Control the VM placement logic in your code.

7
Project 6 Analytical Model for Resource
Allocation

• In this project, I am expecting students to
  derive models of applications and use that to
  determine the amount of resources needed. This
  model could be used by elasticity mechanism to
  determine the amount of additional resources
  needed to satisfy the additional workload.

8
Project 7 Google App Engine application

• Build and deploy a Google App