Contents

1 INTRODUCTION
	1.1 From Mainframes To Personal Computers
	1.2 The Advent of the Information Appliance
	1.3 The Need for Configurable and Modular Operating Systems
	1.4 Performance Implications of Modular Systems
		1.4.1 Potential For Performance Improvements
			1.4.1.1 Code Synthesis
			1.4.1.2 Integrated Layer Processing
			1.4.1.3 PathIDs
			1.4.1.4 Single-Copy TCP/IP
		1.4.2 Implications on Quality-of-Service and Predictability
	1.5 Beyond Modularity: The Path Abstraction
	1.6 Thesis Statement and Contributions
2 PATH ABSTRACTION
	2.1 Communication Network Analogy
	2.2 Path Model
		2.2.1 Basic Path
		2.2.2 Path Processing
		2.2.3 Path Creation
			2.2.3.1 Short Paths
			2.2.3.2 Extending Paths
			2.2.3.3 Optimization Phase
			2.2.3.4 Invariants
			2.2.3.5 Who Creates Paths?
		2.2.4 Generalized Paths
			2.2.4.1 Directionality
			2.2.4.2 Complex Processing
	2.3 Summary and Discussion
		2.3.1 Policy Issues
		2.3.2 Intuitive Models
		2.3.3 Limitations
	2.4 Applications
		2.4.1 Code Optimizations
			2.4.1.1 Code Synthesis
			2.4.1.2 Integrated Layer Processing
			2.4.1.3 PathIDs
			2.4.1.4 Single-Copy TCP/IP
			2.4.1.5 Summary
		2.4.2 Resource Management
			2.4.2.1 Fbufs
			2.4.2.2 Migrating/Distributed Threads
			2.4.2.3 Segregation of Work
	2.5 Related Work
3 SCOUT ARCHITECTURE
	3.1 Overview
		3.1.1 Scout Epochs
	3.2 Modularity
		3.2.1 Module Granularity
		3.2.2 Module Structure
		3.2.3 Module Graph
			3.2.3.1 Compatibility
			3.2.3.2 Runtime Representation and Module Initialization
		3.2.4 Discussion
	3.3 Paths
		3.3.1 Attributes
		3.3.2 Visual Overview of a Scout Path
		3.3.3 Path Object
		3.3.4 Stage
		3.3.5 Interface
		3.3.6 Creation
		3.3.7 Extension
		3.3.8 Optimization
		3.3.9 Destruction
		3.3.10 Evaluation and Discussion
	3.4 Demultiplexing
		3.4.1 Scout Packet Classifier
		3.4.2 The Role of Classifiers
		3.4.3 Realizing the Scout Classifier
			3.4.3.1 Globally Hierarchical Classification
			3.4.3.2 Classification in Non-Demultiplexing Modules
		3.4.4 Evaluation
	3.5 Execution Model
		3.5.1 Thread Scheduling
		3.5.2 Thread Creation
	3.6 Related Work
4 USING PATHS TO OPTIMIZE CODE
	4.1 Preliminaries
		4.1.1 Experimental Testbed
		4.1.2 Test Cases
		4.1.3 Base Case
	4.2 Latency Reducing Techniques
		4.2.1 Outlining
		4.2.2 Cloning
		4.2.3 Path-Inlining
		4.2.4 Last Call Optimization
	4.3 Evaluation
		4.3.1 Test Cases
		4.3.2 End-to-End Results
		4.3.3 Detailed Analysis
			4.3.3.1 Cache Statistics
			4.3.3.2 Processing Time Measurements
			4.3.3.3 Performance Improvement Comparison
			4.3.3.4 Outlining Effectiveness
	4.4 Concluding Remarks
5 USING PATHS FOR RESOURCE MANAGEMENT
	5.1 Building NetTV
		5.1.1 Module Graph
		5.1.2 Paths
			5.1.2.1 Shell Commands
			5.1.2.2 Video Paths
		5.1.3 Base Performance
	5.2 Resource Management
		5.2.1 Queues
		5.2.2 Scheduling
			5.2.2.1 Avoiding Priority Inversion
			5.2.2.2 Scheduling According to Bottleneck Queue
		5.2.3 Admission Control
6 CONCLUSIONS
	6.1 Summary
	6.2 Contributions
	6.3 Future Directions
		6.3.1 CPU Scheduling
		6.3.2 Distributed Paths
		6.3.3 Secure Paths

List of Figures

List of Tables