The ACID Problem in Distributed Systems
What you'll learn: Why keeping ACID guarantees across multiple nodes is fundamentally harder than in a single database.
The Single-Node Comfort Zone
In a traditional single-database system, ACID properties work beautifully. The database engine controls all the data on one machine. When a transaction commits, the database can:
- Atomically update all affected rows in one operation
- Consistently enforce all constraints immediately
- Isolate transactions using local locks
- Durably write to its own disk
Everything happens under one roof, controlled by one manager.
The Distributed Challenge
Now imagine that transaction must update data across three different nodes in different data centers. How do you guarantee that all three nodes commit or all three nodes abort?
- Node A might successfully commit
- Node B might crash mid-transaction
- Node C might lose network connection and not know what happened
Each node runs independently. There's no single authority that can atomically flip all three switches at once.
The Coordination Problem
To maintain Atomicity, you need all nodes to agree on commit or abort. This requires coordination protocols (which you'll learn soon).
To maintain Consistency, constraints spanning multiple nodes need cross-node validation.
This coordination adds:
- Network round trips (latency)
- Failure scenarios (partitions, node crashes)
- Blocking potential (waiting for agreement)
Key Takeaway: ACID guarantees that are simple on a single node become complex distributed coordination problems when data spans multiple nodes, requiring specialized protocols to ensure all-or-nothing outcomes across network boundaries.