CS 603 Description

CS 603 - Advanced Topics in Distributed Systems - (Park, Spring '97)

Introduction (centralized vs. decentralized design).
Networking overview and client/server design.
Group communication (total/causally ordered multicast).
Physical/logical clocks and their application.
Mutual exclusion.
Leader election and fault models (transient/permanent).
Distributed transactions.
Concurrency control (locking, optimistic, time-stamp).
Distributed deadlock detection/prevention.
Theoretical aspects of consensus problems (some impossibility results).
Advanced fault models and their relations (worst-case, k-sparse, Byzantine).
Fault-tolerant distributed computing (space/time/information redundancy, hierarchical, resiliency).
Load balancing (static/dynamic optimization, on-line algorithms and competitive ratio, process migration, implementation mechanics).
Noncooperative vs. cooperative resource allocation algorithms (system/Pareto optimality, Nash equilibria, computational markets).
Distributed scheduling (optimality/fairness, quality of service).
Real-time scheduling (RMA, EDF, WFQ, Solaris implementation issues).
Name resolution, access patterns, and optimal name server construction.
Distributed caching (cache consistency, replica management).
Distributed file systems (file system properties, optimal file allocation, heavy-tailed file size distributions and aggregate behavior).
Distributed shared memory (sequential, causal, PRAM, weak, release, and entry consistency; implementations and performance).
Future trends.