Short note: Distributed system implementation

A distributed system runs its objects on several computers connected by a network. Implementing it raises issues that do not appear in a single-machine program:

1. Network latency and bandwidth: remote calls are much slower than local calls, so "chatty" designs with many small calls must be avoided.
2. Partial failure and reliability: a node, link or message can fail while the rest of the system keeps running; timeouts, retries and recovery are needed.
3. Concurrency: many clients use shared objects at the same time, so locks, transactions and synchronisation are needed to prevent inconsistent data and deadlock.
4. Data consistency and replication: copies of data on several nodes must be kept consistent.
5. Heterogeneity: nodes may use different hardware, operating systems and languages, so standard protocols and middleware (CORBA, RMI, web services, REST) are needed.
6. Location transparency and object naming: objects must be located and referenced across machines (using a naming or registry service).
7. Security: authentication, authorisation, and encryption of data sent over the network.
8. Scalability and load balancing: the system should keep working well as the number of users and nodes increases.
9. Object passing: parameters must be marshalled (serialised) to be sent between nodes.
10. Testing, debugging and deployment: these are harder, because errors can depend on timing and on how the system is distributed.

Design patterns such as Proxy, Broker and Facade, together with middleware, help to manage these issues.