Distributed Computing Systems CSCI 47806780 Scalability Scalability with

Distributed Computing Systems CSCI 4780/6780

Scalability • Scalability with respect to size • Scalability with respect to geographical location • Scalability with respect to management Concept Example Centralized services A single server for all users Centralized data A single on-line telephone book Centralized algorithms Doing routing based on complete information Scalability Problems

Example • Problem: Design a Highly Scalable Map (Direction) Service • Data – Route information • Task – Computing shortest (least cost) routes • Option 1: Store all data at a single server, perform all computations there • Option 2: Store data centrally, but perform operations on multiple nodes • Option 3: Store data at multiple locations and perform operations in a distributed fashion

Characteristics of Distributed Algorithms • Single machine does not have complete information • Decisions are made based on local information • Failure of a node does not cause the algorithm to fail • No global clock • Distributed algorithms are much harder than centralized algorithms

Geographical Scalability Challenges • Synchronous communication – Waiting for a reply does not scale well!! • Unreliable networks – Broadcasting does not work • Centralized components – Will eventually become bottlenecks • Multiple administrative domains – Security mechanisms do not come for free

Scalability Techniques • Hiding communication latencies – Asynchronous communication – Reduce amount of data transmitted • Distribution – Spreading work across system • Caching and replication – Make copies of data and services • Balance load – Avoid hot spots

Scaling Techniques 1. 4 The difference between letting: a) a server or b) a client check forms as they are being filled

Scaling Techniques (2) 1. 5 An example of dividing the DNS name space into zones.

Pitfalls • • Network is reliable Network is secure Network is homogeneous Topology does not change Latency is zero Bandwidth is infinite Transport cost is zero Single administrator

Types of Distributed Systems • High-Performance Computing Systems • Cluster computing • Grid computing • Distributed Information Systems • Transaction processing systems • Enterprise application integration • P 2 P systems • Distributed Pervasive Systems • Mobile networks • Sensor networks

Cluster Computing • • • Super computing using PCs/workstations Used for parallel programming Homogeneity among nodes Master allocates nodes to tasks Compute nodes run standard operating systems (and may be a very thin middleware) • Middleware consists of parallel programming libraries

Cluster Computing Systems An example of a cluster computing system.