Abstraction

1. Why Are Abstractions Important?
Abstractions simplify complex systems by hiding lower-level implementation details. They:

Allow developers to focus on solving higher-level problems.

Enable modularity: changes in one layer don’t affect others.

Promote reuse, scalability, and testability.

Are foundational in both OS (e.g., file systems) and distributed systems (e.g., databases, RPC).

2. Network Abstractions: Remote Procedure Calls (RPC)
RPC enables communication between services across a network, hiding the complexity of:

Data serialization/deserialization

Network communication protocols

Retry logic and error handling

Advantages:

Makes remote calls look like local method calls.

Reduces boilerplate for service-to-service interactions.

Popular frameworks: gRPC, Thrift, Apache Avro.

Caveats:

Network issues can introduce latency/failures.

RPC is not the same as function calls: developers must account for distributed system challenges (timeouts, retries, etc.).

3. Spectrum of Consistency Models
Consistency in distributed systems balances availability, latency, and correctness.

Models include:

Strong Consistency – Reads always reflect latest writes. Simplifies reasoning but affects availability.

Eventual Consistency – All replicas eventually converge. Faster but may show stale data.

Causal Consistency – Preserves cause-effect relationships.

Read-Your-Writes / Monotonic Reads – Guarantees that a user's session observes its own updates.

Tradeoffs: CAP theorem dictates that in network partitions, you must choose between consistency and availability.

4. The Spectrum of Failure Models
Failures are inevitable in distributed systems. Types include:

Crash Failures – Server stops working (e.g., power loss).

Omission Failures – Messages dropped or not sent.

Timing Failures – Response not within expected time.

Byzantine Failures – Arbitrary/malicious behavior.

Key takeaways:

Design for failure: Use timeouts, retries, replication, and monitoring.

Systems must tolerate partial failures without breaking completely.

Consensus protocols (e.g., Paxos, Raft) handle agreement in unreliable environments.