Abstractions

Question 1

Why do we need abstraction?

Answer 1

It helps us see the big picture why obsfucating details that we don’t need

Question 2

What is a database abstraction called?

Answer 2

Transaction
Mutate from one consistent state to another, only providing info on success or failure

Question 3

What is example of a network abstraction?

Answer 3

Remote procedure calls (RPC)

Question 4

What are consistency models?

Answer 4

Abstractions to reason about correctness of a distributed system doing concurrent data reads, writes and mutations

Question 5

What are the four types of consistency models, from weakest to strongest?

Answer 5

Eventual consistency, causal consistency, sequential consistency, strict consistency/linearizability

Question 6

What is the CAP or Brewer”s theorem?

Answer 6

Can only guarantee two out of three
1. Consistency
2. Availability
3. Partition tolerance

Question 7

How many types of failure models are there?

Answer 7

Five

Question 8

Name the failure models, from easy to deal to difficult.

Answer 8

1. Fail stop - halts permanently but other nodes can still communicate
2. Crash - half’s silently and other nodes cannot detect this
3. Omission - node fails to send or receive messages. Send omission failure (fails to respond to incoming request) and receive omission failure
4. Temporal - generate results but too late to be useful, usually from bad algo design or sync between clocks
5. Byzantine - random behavior, bug or malicious attacks