Storage service providers (SSPs) are fault-prone black boxes operated by a third party. Prudent clients should avoid strong assumptions about the integrity of data stored remotely and implement some form of end-to-end checks. Based on these premises, this paper describes in detail a protocol, layered on top of most cloud storage services, that enforces fork-join-causal (FJC) consistency to guarantee data integrity. The protocol, dubbed Depot, increases reliability by replicating both data and metadata associated with update history.

The paper is well organized and explains Depot’s architecture, operation, protocol, consistency model, and implementation. It also briefly evaluates performance of a prototype and compares its cost with other simpler protocols. This exercise is geared toward evaluating the cost of trust. The authors conclude that, while Depot’s read operations do not cost much, writes (updates) can be quite costly and add more than 50 percent to baseline central processing unit (CPU) consumption and more than 30 percent for remote storage space.

There is also an interesting discussion on related work in which the authors classify existing systems in terms of high availability and fault tolerance.