QoS Negotiation in Real-Time No critique for Mondays class Systems - - PDF document

1

Announcements

Motes and programming boards available No critique for Monday’s class

QoS Negotiation in Real-Time Systems

By Stephanie McCarthy

Overview

Motivation QoS negotiation model RTPOOL Implementation Evaluation and future work

QoS Negotiation in Real-Time Systems Definitions

What is QoS? What is automated flight control?

QoS Negotiation in Real-time Systems Motivation

QoS negotiation vs. Traditional method Allow graceful QoS degradation Predictability Used for long-lived real-time applications

QoS Negotiation Real-Time Systems QoS model

QoS levels, rewards, rejection penalty

Reward is degree of satisfaction to be achieved from the QoS level Rejection penalty is the penalty that incurs to application if the request is rejected

Client specifies set of negotiation options and penalty of rejecting request Each option consists of a QoS level and reward value for client to receive from provider

2

QoS Negotiation in Real-Time Systems Qos Model

Client Service Provider

Request

1) Negotiation Option a) Acceptable QoS levels b) Reward 2) Rejection Penalty

QoS Negotiation in Real-Time Systems QoS model

QoS levels are expressed in terms of parameters Examples of parameters

Traffic delay and jitter requirements in real-time communications Required processor bandwidth in processor capacity reservation

QoS Negotiation Real-Time Systems QoS Model

Guaranteeing a clients request is the certification of the request to receive service at one of the QoS levels but the service provides picks the actual QoS level The service provider can switch the QoS level if it perceives increased utility.

QoS Negotiation Real-Time Systems QoS model

Advantages of letting the service provider pick the QoS level

Application code makes no assumption of resource availability or capacity Achieve higher over-all performance while maintaining QoS guarantee

QoS Negotiation Real-Time Systems RTPOOL

Designed to test QoS model Pool of shared processors Provide timeliness guarantees for flight control Clients are application tasks

QoS Negotiation in Real-Time Systems RTPOOL

A clients task is specified by parameters Task execution time should not be included in parameter Programmer can define different versions

• f the same task

3

QoS Negotiation in Real-Time Systems RTPOOL

Local Optimization Heuristic

Re-computes set of QoS levels when new task arrives

Task Rejection Unfulfilled potential reward (UPR) Load Sharing Algorithm

Implements a distributed QoS-optimization protocol

QoS Negotiation in Real-Time Systems Implementation

Scheduling and QoS negotiation

Qthread Lowest layer of RTPOOL Force_Negotation() primitive initiate QoS negotiation

QoS Negotiation in Real-Time Systems Implementation

Invocation Migration

Implements QoS optimization protocol QoS optimization Heuristic

Force_negotiation() Wire_task()

QoS Negotiation Real-Time Systems Implementation

Pool Membership API

Maintain consistent view of current membership in the shared resource pool Subscribe_to_pool()

Unfulfilled potential reward Machine failure

QoS Negotiation in Real-Time Systems Implementation

Communication API

Executable applications composed of tasks Send() and Receive() primitives Communication protocol stack implemented using a xKernal 3.2 and layered on top of UDP/IP stack

QoS Negotiation in Real-Time Systems Application to Flight Control

QoS negotiation supports mission planning techniques in context of CIRCA CIRCA-build a set of control plans to keep system safe Probabilistic planning Mission Planner

4

QoS Negotiation in Real-Time Systems Description of Flight Tasks

Guid Ctrl FNav SNav

QoS Negotiation in Real-Time Systems Evaluation

Heuristic testing Aircraft performance with ctrl and Fnav tasks Load Sharing

QoS Negotiation in Real-Time Systems Evaluation

Task is fixed decreased CPU speed (increases task execution time) Optimality is compromised but graceful degradation is achieved

QoS Negotiation in Real-Time Systems Evaluation

Tested on ability to control the simulated F-16 during flight Evaluation using Ctrl and Fnav tasks which are coupled

QoS Negotiation in Real-Time Systems Evaluation

Two Machines run functionally Shut down one machine to simulate failure just after take-off Conventional method failure

QoS Negotiation in Real-Time Systems Future Work

Currently studying alternative QoS

• ptimization methodologies and scalability
• f this QoS approach

Negotiable fault tolerance QoS as an extension to RTPOOL