1 / 16

Project Design

Project Design. Alain Esteva-Ramirez School of Computing and Information Sciences Florida International University. Bárbara Morales-Quiñones Department of Computer Engineering University of Puerto Rico-Mayaguez. REU Summer Program. 07/17/2007. Outline. Recap of Previous Phase

nadda
Télécharger la présentation

Project Design

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Project Design Alain Esteva-Ramirez School of Computing and Information Sciences Florida International University Bárbara Morales-Quiñones Department of Computer Engineering University of Puerto Rico-Mayaguez REU Summer Program 07/17/2007

  2. Outline • Recap of Previous Phase • Introduction • What are we doing now? • How the system will work • Request Scheduling • Diagrams • Algorithm outline • Goals • Questions?

  3. Recap Where we left off last time… • Safe Adaptation with Validation Strategy to Testing AC Systems at Runtime • Performed background on safe adaptation (Zhang) • Focused on how to test self-optimization changes • Surveyed tools to support testing adaptive AC systems • Tools Survey • Looked at <number> of automated tools • Dependency Analysis: JDepend and Cap • Performance Metrics: TPTP • Unit Testing Support: Cobertura and JUnitPerf

  4. Introduction What are we doing now? • Problem • Design an adaptable autonomic container. • Incorporate self-optimization capability • Apply container in the context of a real problem • Integrate tools to support validation of self-optimization • Solution • Autonomic Request Scheduling Container

  5. How the system will work Autonomic Request Scheduling • Recall Stevens and Parsons autonomic stack of random numbers (REU 2006) • Stack self-configured by increasing size • Container will now hold user requests that need to be serviced. • Agents dequeue requests according to some high-level scheduling algorithm, and service them. • Self-optimization changes strategy in attempt to prevent starvation

  6. How the system will work Request Scheduling Strategies • First come, first serve (FCFS) • Decision mode: non-preemptive • Comments: A potential problem is when a short request follows a long one. • Shortest Request Next (SRN) • Decision mode: non-preemptive • Comments: SRN is better than FCFS for short request, but long request may have to wait for some time for service. To avoid this problem, a request's time limit can be used.

  7. Design Constraints • Need a flexible design to facilitate seamless integration of new components • E.g. New scheduling strategies • High-level design of autonomic system must conform with standard architecture of autonomic computing (IBM ‘06] • Self-testing framework must conform with safe adaptation with validation strategy (King et al.) • Reuse as much as possible, Patterns

  8. High Level Architecture

  9. Generic Manager Design

  10. Minimal Class Diagram

  11. Selected Design Patterns • Façade (Interface pattern) • Provides an interface that makes a subsystem easy to use • Strategy (Operation pattern) • Encapsulates alternate strategies in separate classes • Observer (Responsibility pattern) • Define a dependency between objects • When one changes state, dependents are notified and updated automatically

  12. Algorithm Outline Strategy swap : FCFS to SRN • Precondition: Three agents and approximately 15 requests • FCFS is the current strategy. • Agent3 is at finished, the others are working on heavy requests. • Steps: • The Monitor of the TAMSelfOptim through RATouchData knows that the AgentPool has the third agent free. (1) • The Monitor of the TAMSelfConfig through RCTouchData knows that the next request in the RequestContainer is heavy.(2) • Both managers store that information in the Knowledge Source through the KSFacade. • The OAMReqSched analyses that information (1 and 2), thus changes the strategy in the SchedStratey to SRN. • Then the OAMReqShed notifies the TAMSelfOptim through the KSFacade that the third agent should be set to ready.

  13. Activity Diagram • TAMSelfOptim detects that the Agent Pool has a free agent. • TAMSelfConfig detects that the next request in the Request Container is heavy. • Both managers write this state information to the Knowledge Source (KS). • OAMReqSched analyzes information from KS and detects that more than 60% of other agents are working on heavy requests. • OAMReqSched requests to swap strategy to SRN. • OAMReqSched notifies that a free agent is now ready.

  14. Activity Diagram Strategy swap : FCFS to SRN

  15. Goals • Have a solid adaptive framework • Test it during safe ad • Write a technical paper • Write a student paper

  16. Questions? Questions, comments and queries.

More Related