Distributed Systems and Algorithms Sukumar Ghosh University of Iowa Fall 2009

What is a distributed system? 5 2 1 4 8 3 7 10 6 9 A network of processes. The nodes are processes, and the edges are communication channels.

Fact Processor population is exploding. Technology has dramatically reduced the cost of processors, and user demands for services have increased the scale of systems.

Examples Large networks are very commonplace these days. Think of the world wide web. Other examples are: - Social Networks Sensor networks BitTorrent for downloading video / audio Skype for making free audio and video communication Process control networks Computational grids (OSG, Teragrid, SETI@home) Network of mobile robots

Sensor Network Checking the structural integrity

Mobile robots I-Swarm Robot (See a YouTube video on the I-Swarm Robots)

Goal P Q Typically, the application layers of the processes interact with one another to meet a common objective, or achieve individual objectives using shared resources. Copy X from P to Q (P&Q may be two processes 10,000 miles apart)

Goal continued Distributed computing relies on inter-process communication, which involves other layers of networking too (like TCP implementing a reliable end-to-end communication channel, or the Media Access protocol used in Ethernet LAN or Wireless networks). What are the distributed computing issues in establishing inter-process communication? P Q Create a reliable channel from P to Q

Why distributed systems Geographic distribution of processes Resource sharing (example: P2P networks, grids) Computation speed up (as in a grid) Fault tolerance

Important services Internet auction Web search Google earth Internet banking Web search Net meeting Distance education Video distribution Internet auction Google earth Skype Publish subscribe

Important issues Clocks are not synchronized No shared address space Knowledge is local Clocks are not synchronized No shared address space Topology and routing : everything is dynamic Scalability: all solutions do not scale well Processes and links fail: Fault tolerance

Some common subproblems Leader election Mutual exclusion Time synchronization Distributed snapshot Replica management Consensus

Implementation Practical distributed systems must have a real network as its backbone. However, such systems can be simulated on a shared-memory multiprocessor, or even on a single processor. (How will you do it?)

Models We will reason about distributed systems using models. There are many dimensions of variability in distributed systems. Examples: types of processors interprocess communication mechanisms, timing assumptions failure classes, security features, etc

Models Models are simple abstractions that help overcome the variability -- abstractions that preserve the essential features, but hide the implementation details from observers who view the system at a higher level. We don’t care whether a process runs on a PC or a Mac. A link may use optical communication or radio communication For simplicity, in a specific model, we may assume that all clocks are perfectly synchronized.

A classification Client-server model Server is the coordinator Peer-to-peer model No unique coordinator

Parallel vs Distributed In both parallel and distributed systems, the events are partially ordered. In parallel systems, the primarily issue is speed-up In distributed systems the primary issues are fault-tolerance, availability, and scalability

Objective of the course To understand how a system works, why it fails, how can we guarantee our design, so that it behaves the way we want it to behave. A system that “sometimes works” is no good. We are not discussing implementation or programming, although they are also important issues.