Scenarios

_Below are several scenarios that will be investigated by CERAS researchers_

Scenario 1. High Adaptation (flexibility)

At 11:00 a.m. on July 10, Susan, the head of Research at Toronto Hospital, is told that her department has to help with prognosis the spread of a pandemic disease that has just hit Toronto. Susan realizes that she can use her Patient Analyis program. But she needs to add ten more columns to the UI to capture the new type of data from the new admitted patients and she also needs to integrate this new Patient Analysis program with a new prognosis algorithm, Predict, recently announced by University of Toronto researchers. With the current IT situation, Susan will ask the IT department to make the changes to the Patient Analysis program and to implement and deploy the new Predict algorithm. It will take several months to implement, test, buy the required hardware and deploy the new system. In an Adaptive system, Susan will most likely change the Patient Analysis program herself by adding the required columns because the system is highly adaptable and she can have it ready by 11:10am, the same day, so she can start collecting data for the new admitted patients. For the new Predict algorithm which requires huge processing power and integration with the Patient Analysis program, she will delegate it to Nancy, from the IT department. Nancy will quickly model the program, test it and deploy it in the adaptive virtual infrastructure that the Toronto Hospital shares with the partner universities and hospitals. The new program will be fully functional in a matter of days.

Scenario 2. Self-optimization scenario

Preconditions:

• The applications A, B, C use processing power and storage according to an workload profile
• There are resources available in the system

Trigger: the application A experiences a surge in the workload << intensity- arrival rate, or mix-change in type of transactions>> Normal Flows:

• The Virtual computing Infrastructure (VCI) layer observes the change in the workload
• VCI attempts a local optimization (on the local cluster)
• Tune the existing application (like DB2, middle tier)
• Reallocates the local resources (more resources to the database and less resources to the application servers, etc..)
• VCI attempts a global optimization (across clusters)
• Find available resources in other clusters
• Migrate parts of the application A to other clusters
• Initialize and run the parts of A on other clusters

Exceptional Flows

• the optimization attempt fails, a rollback is needed

Postconditions: a) the performance parameters are within acceptable limits; b) the application is in a controllable state

Scenario 3. Self-healing scenario

Preconditions:

• The application A is running
• There are resources available in the system

Trigger: the application A experiences malfunction symptoms << lack of correlation between workload intensity - arrival rate and response time; excessive memory consumption, failed transactions>> Normal Flows:

• The Virtual computing Infrastructure(VCI) notices the symptoms
• VCI discover the cause (a faulty application server)
• By applying method X
• By applying method B
• VCI attempts to fix the problem
• Stop the faulty server
• Find available resources in other clusters
• Migrate parts of the application A to other clusters
• Initialize and run the parts of A on other clusters

Exceptional Flows

• the recovery fails, a rollback is needed

Scenario 4. Virtual Computing Lab

Preconditions:

• Database of Lab images available
• Resources across multiple sites are virtualized

Trigger: Instructor X requires Y machines for the Lab L Normal Flows:

• The Virtual computing Infrastructure(VCI) layer receives a request for Y machines
• VCI find Y available resources
• VCI copies the image of the Lab L on those machines
• VCI makes the IP addresses of the Labs available
• VCI updates the resource usage tables