Distributed Systems

https://aws.amazon.com/builders-library/challenges-with-distributed-systems/?did=ba_card&trk=ba_card

Types of Distributed Systems

• Offline - These are systems that can tolerate being offline. These systems are not latency sensitive. These include batch processing systems, big data analysis clusters, movie scene rendering farms, protein folding clusters, and the like. While far from trivial to implement, offline distributed systems get almost all of the benefits of distributed computing (scalability and fault tolerance) and almost none of the downsides (complex failure modes and non-determinism).
• soft real-time systems - These are critical systems that must continually produce or update results, but they have a relatively generous time window in which to do so. These systems can go offline for minutes or hours with no customer impact.
• hard real-time distributed systems - These are request/reply services. What makes them difficult is that requests arrive unpredictably and responses must be given rapidly

System Models

Distributed System Models & Distributed Systems Theory.

https://www.scylladb.com/glossary/consistency-models/

https://jepsen.io/consistency

https://muratbuffalo.blogspot.com/2024/10/ddia-chp-7-transactions-part-1.html

The definition of consistency in distributed systems is overloaded and different than its definition in ACID. In ACID, consistency means the database doesn’t violate any explicit constraints (foreign key relationships, etc). However, in distributed systems, consistency includes the atomicity, availability (staleness), and isolation of operations.

In many context, we talk about ‘strong consistency’ and ‘weak/eventual consistency’. However, these are not actually defined consistency models, and therefore can be confusing. For example, a system may be said to be strongly consistency, but not implement isolation between process, which therefore is not ‘serializable’ (which most people expect when we mean strong consistency).

Isolation Faults