By Krishnamurthy et Al. Presented by David Girsault Integration of QoS-enabled Distributed Object Computing Middleware for Developing Next-Generation Distributed Applications By Krishnamurthy et Al. Presented by David Girsault
Introduction Distributed frameworks Beginning => improve inter-operability Now => demand for better QoS What needs for resource management? Before run-time => already exists During run-time => research
Overview of the Adaptive QuO-Middleware QuO = Quality Objects Allows specifying QoS requirements Control and monitoring information Behavior for dealing with run-time problems RPC contract + QoS guaranty Middleware in charge of QoS respect
Unified Middleware-Centric Qos API Many != QoS frameworks Standardization necessary Need for an API for QoS Separation between application and protocol Translate QoS parameters to the network Notification system for environment change
ACE QoS API Two implementations: Windows, Unix Extensibility: easy to add network protocols Portability: easy to modify for a platform Event Notification: when QoS change QoS capabilities: resource reservation system
Application to the CORBA Audio/Video Streaming Service QoS parameters: Application class (interactive or not) Media types (audio, video…) Adaptation policies (coeff. of compression) CORBA A/V streaming service: Flexible and portable CORBA Efficient low-level transport
The TAO A/V Streaming Service QoS Framework QoS Mapping: translation between network-level and application-level QoS Monitoring: application adaptation to resources at run-time QoS-Based transport API: Guarantees required bandwidth
Shipboard Dissemination of Unmanned Air Vehicle Video MPEG data path: UAV Onboard distribution center Video display host
Adaptation and Performance Under Load Adaptation vs Non-adaptation Testing overloading distribution center A/V Streams vs Ad-Hoc TCP Results: similar performance Time latency irregular with A/V streams
Conclusion QoS-enabled middleware: Resource variation at run time Decreases time and effort for QoS requirements Increases modularity and portability Resource variation at run time System response improvement Better service to end users