You are here: ProjetSEG Web>CsmExport>CsmRulesForLqnGenerator (30 Mar 2007)

CSM Rules for Input to the CSM2 LQN? Generator

Originally proposed by Dorin B. Petriu, Ph.D. candidate, Carleton University (March 30, 2007)

Basic Rules

  1. All Scenarios must begin with a Start path connection
  2. All top-level Scenarios must begin with a Start path connection that includes a Workload
  3. Conversely, all Scenarios that begin with a Start path connection with a Workload are treated as top-level Scenarios
    • so a sub-scenario with a Workload will be traversed every time it is included in a complex Step as well as once as a stand-alone top-level Scenario, sub-scenarios that should not be traversed on their own should not have any Workload attached
  4. All CSM Scenarios must finish with an End path connection
  5. All Steps (and Resource Acquire/Release) must be connected to predecessor and successor path connections mutliple Steps without
    • DA: Not sure about the last part
  6. Conversely, all path connections must be connected to source and target Steps (or Resource Acquire/Release). However:
    1. Start path connections cannot have source Steps and must have 1 target Step
    2. End path connections must have 1 source Step and cannot have target Steps
    3. Sequence path connections must have 1 source Step and 1 target Step
    4. Fork and Branch path connections must have 1 source Step and 2 or more target Steps
    5. Join and Merge path connections must have 2 or more source Steps and 1 target Step
  7. Mutliple Steps without path connections are not allowed
  8. Multiple path connections without Steps are not allowed
    • DA: For the two above, you mean sequences of Steps without path connections and vice-versa? This is redundant with the previous info. Also, this is already enforced by the XSD schema
  9. All Resource Acquire/Release elements must be contained in the Scenario
  10. Resource Acquire/Release elements must not be contained inside Steps, even though the schema allows for them to be so contained.
  11. All resources that are not ProcessingResources must be acquired/released explicitly with Resource Acquire/Release elements
  12. Active Components must have a host relationship with a ProcessingResource
    1. The host ProcessingResource for a Component is implicitly acquired when the Component is acquired
    2. The host ProcessingResource for a Component is implicitly released when the Component is released
    3. ProcessingResources must not be explicitly acquired/released
  13. Complex Steps with sub-Scenarios with multiple Start/End path connections must have Input and Output Bindings to indicate which Start/End path connections are connected to the main flow of the path

Advanced Rules

  1. All Steps should have a hostDemand attribute
  2. The target Steps of a Branch path connection should have a probability
  3. That probability is the probability for that particular branch
  4. It is preferable that Steps have a component attribute that points to the Component in whose context the Step executes
    • DA: Otherwise?
  5. The number of resourceUnits acquired over the traversal of a top-level Scenario must equal the number of resourceUnits released through the traversal of a top-level Scenario
    • DA: Otherwise? This might be difficult to enforce at the UCM level. Is this detected by the tool and does this prevent the LQN from being generated?
  6. It is not permissible at any point in the traversal of a Scenario to release more units of a resource than were previously acquired along that Scenario
    • DA: ditto
  7. The End path connection of a Scenario with a ClosedWorkload must occur in the same Component execution context as the Start path connection
    • DA: Why?

Attributes and Default Values

  1. There are no defaults for Workload attributes
  2. The Step hostDemand attribute has a default value of 0
  3. The Step probability attribute has a default value of 1
  4. The Step repCount attribute has a default value of 1
  • DA: Any defaults for RA/RR quantities, for component multiplicities...

Known Limitations of the CSM2 LQN? Generator

  1. Does not properly handle sequences of successive Resource Acquire/Release elements that do not have at least one execution Step in between them
  2. Need to insert a "dummy" execution Step (with hostDemand of 0) between any sequence of successive Resource Acquire/Release elements
  3. Does not properly handle Fork-Join or Branch-Merge constructs where the Fork/Branch and Join/Merge occur in the execution context of different Components
  4. Does not handle at all Scenarios where Join/Merge occurs before Fork/Branch

Concepts Still To Be Resolved

  1. How are resources used after a Fork?
  2. Assume that all parallel branches of a Fork share the resources held at the Fork?
  3. How to interpret unmatched ResourceReleases along a parallel branch?
  4. Assume that the resource is released for all the parallel branches?
  5. At what point along the other parallel branches do they lose the use of the resource?
  6. What is the meaning of resources passed at a Fork?
    1. Assume that the parallel branch which get passed a resource owns it exclusively?
  7. What about all the other resources?
    1. Does the parallel branch own any of them?
    2. Does the parallel branch own only the resources it is passed?
    3. Does the parallel branch share all resources that have not been explicitly passed to it or any other parallel branch?

-- Daniel Amyot - 30 Mar 2007

Topic revision: r1 - 30 Mar 2007 - 15:14:11 - Daniel Amyot
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback