Goals and design challenges of DS

Goals of a distributed system

A distributed system is a collection of independent computers, connected by a network and running distribution middleware, that appears to its users as a single coherent system. Its components communicate and coordinate their actions only by passing messages (Tanenbaum; Coulouris).

• Resource sharing: users and programs share printers, files, data, storage and compute power across the network (for example the web, cloud storage).
• Transparency: hide the fact that processes and resources are physically spread over many machines.
• Openness: the system offers services through standard, published interfaces so that parts from different vendors can interoperate and be extended.
• Scalability: the system keeps working well as users, resources and geographic spread grow.
• Concurrency and fault tolerance: many users work at once, and the system keeps running when some component fails.

Design challenges

1. Heterogeneity: different networks, hardware, operating systems, programming languages and data formats must work together. Middleware (CORBA, Java RMI) and standards hide the differences.
2. Openness: interfaces must be published, specified (for example by IDL) and extensible without breaking existing parts.
3. Security: confidentiality, integrity and availability must be protected over open networks. Encryption, authentication and access control are needed, and denial-of-service attacks are hard to stop.
4. Scalability: the design must avoid central bottlenecks (single server, central table, central algorithm). Techniques are replication, caching, partitioning and distribution of work.
5. Failure handling: components and links fail independently and partially. We need failure detection, masking (retransmission, replication), tolerance and recovery.
6. Concurrency: shared resources are accessed at the same time, so synchronisation and consistency control are needed.
7. Transparency: hiding distribution completely is costly and sometimes impossible (network latency, partitions).
8. Quality of service and time: timeliness and reliability must be guaranteed, and the lack of a global clock makes coordination difficult.