Foreword
Preface
DAIS 2023 Organization
Contents
Distributed Algorithms andΒ Systems
TADA: A Toolkit for Approximate Distributed Agreement
1 Introduction
2 Related Work
3 Approximate Distributed Agreement
3.1 Mean-Subsequence-Reduce (MSR)
4 A Template for Approximate Agreement
5 How to Use the Toolkit
5.1 Specifying Generic Primitive Variables
6 Evaluation
7 Conclusion
References
Studying the Workload of a Fully Decentralized Web3 System: IPFS
1 Introduction
2 IPFS
3 Methodology
3.1 Processing the Requests
3.2 Locating the Content Providers
3.3 Analyzing the Data
3.4 Implementation Details
4 Results
4.1 Requests
4.2 Providers
4.3 Requested Content Vs. Provided Content
5 Related Work
6 Conclusion
References
Community-Based Gossip Algorithm for Distributed Averaging
1 Introduction
2 Background
2.1 Distributed Averaging
2.2 Gossip Algorithms
3 Previous Work
4 Community-Based Gossip Protocol
5 Experimental Methodology
5.1 Metrics
5.2 Networks
5.3 Simulator
6 Results and Discussion
6.1 Effect of Modularity on Distributed Averaging Performance
6.2 Predictive Value of Modularity Metrics
6.3 Structural and Functional Properties of Boundary Nodes
6.4 Performance of Community-Based Gossip
7 Conclusions
References
Data Management
Transactional Causal Consistent Microservices Simulator
1 Introduction
2 Related Work
3 Running Example
4 Simulator
4.1 Achieving Transactional Causal Consistency
4.2 Functionality Execution
4.3 Event Handling
5 Architecture
6 Evaluation
6.1 Threats to Validity
7 Conclusion
References
The Impact of Importance-Aware Dataset Partitioning on Data-Parallel Training of Deep Neural Networks
1 Introduction
2 Background and Related Work
2.1 DNN Data-Parallel Training (DPT)
2.2 Prior Work on Example Importance
3 Importance-Aware DPT
3.1 Warmup Training
3.2 Importance Calculation
3.3 Dataset Partitioning Heuristics
3.4 Intervals of Model Training
4 Implementation in PyTorch
4.1 Importance Calculation
4.2 Dataset Partitioning Heuristics
4.3 Modified Training Loop for Importance-Aware Training
5 Evaluation
5.1 Experimental Setup
5.2 Different Dataset Complexities
5.3 Different Models
5.4 Different Partitioning Heuristics
5.5 Different Importance Metrics
5.6 Added Overheads
6 Conclusion
References
Distributed Architectures
Runtime Load-Shifting of Distributed Controllers Across Networked Devices
1 Introduction
2 Problem Statement
2.1 Analogies with Systems in the Literature and in the Industry
3 Proposed Architecture
3.1 Load-Shifting Spectrum: An Example
3.2 Limitations and Technological Constraints
4 Proof of Concept
4.1 Technology Selection
4.2 System to be Controlled: F
4.3 Common Data Model: M
4.4 Monitor/Controller: N
4.5 Renderer: H
4.6 Final Design
5 Evaluation
5.1 Test Environment and Qualitative Assessment
5.2 Performance Evaluation
6 Conclusion and Future Work
References
EdgeEmu - Emulator for Android Edge Devices
1 Introduction
2 Related Work
2.1 Network Simulators/Emulators
2.2 Fog, Edge and Cloud Simulators
2.3 Test Frameworks
3 Android Virtual Devices and Networking
4 EdgeEmu Emulation
5 Implementation
6 Evaluation
6.1 Testing Environment
6.2 Ping Test
6.3 File Sharing Test
6.4 Number of Emulators
7 Conclusion
References
Author Index