Search

Atomic shared objects, whose operations take place instantaneously, are a powerful technique for designing complex concurrent programs. Since they are not always available, they are typically substituted with software implementations. A prominent condition relating these implementations to their atomic specifications is linearizability, which preserves safety properties of programs using them. However linearizability does not preserve hyper-properties, which include probabilistic guarantees about randomized programs. A more restrictive property, strong linearizability, does preserve hyper-properties but it is impossible to achieve in many situations. In particular, we show that there are no strongly linearizable implementations of multi-writer registers or snapshot objects in message-passing systems. On the other hand, we show that a wide class of linearizable implementations, including well-known ones for registers and snapshots, can be modified to approximate the probabilistic guarantees of randomized programs when using atomic objects. This is joint work with Hagit Attiya and Constantin Enea.

A key way to construct complex distributed systems is through modular composition of linearizable concurrent objects. A prominent example is shared registers, which have crash-tolerant implementations on top of message-passing systems, allowing the advantages of shared memory to carry over to message-passing. Yet linearizable registers do not always behave properly when used inside randomized programs. A strengthening of linearizability, called strong linearizability, has been shown to preserve probabilistic behavior, as well as other "hypersafety" properties. In order to exploit composition and abstraction in message-passing systems, it is crucial to know whether there exist strongly-linearizable implementations of registers in message-passing. This paper answers the question in the negative: there are no strongly-linearizable fault-tolerant message-passing implementations of multi-writer registers, max-registers, snapshots or counters. This result is proved by reduction from the corresponding result by Helmi et al. The reduction is a novel extension of the BG simulation that connects shared-memory and message-passing, supports long-lived objects, and preserves strong linearizability. The main technical challenge arises from the discrepancy between the potentially minuscule fraction of failures to be tolerated in the simulated message-passing algorithm and the large fraction of failures that can afflict the simulating shared-memory system. The reduction is general and can be viewed as the inverse of the ABD simulation of shared memory in message-passing.

An elegant strategy for proving impossibility results in distributed computing was introduced in the celebrated FLP consensus impossibility proof. This strategy is local in nature as at each stage, one configuration of a hypothetical protocol for consensus is considered, together with future valencies of possible extensions. This proof strategy has been used in numerous situations related to consensus, leading one to wonder why it has not been used in impossibility results of two other well-known tasks: set agreement and renaming. This paper provides an explanation of why impossibility proofs of these tasks have been of a global nature. It shows that a protocol can always solve such tasks locally, in the following sense. Given a configuration and all its future valencies, if a single successor configuration is selected, then the protocol can reveal all decisions in this branch of executions, satisfying the task specification. This result is shown for both set agreement and renaming, implying that there are no local impossibility proofs for these tasks.

Easy proofs are given, of the impossibility of solving several consensus problems (Byzantine agreement, weak agreement, Byzantine firing squad, approximate agreement and clock synchronization) in certain communication graphs. It is shown that, in the presence of m faults, no solution to these problems exists for communication graphs with fewer than 3m+1 nodes or less than 2m+1 connectivity. While some of these results had previously been proved, the new proofs are much simpler, provide considerably more insight, apply to more general needs of computation, and (particularly in the case of clock synchronization) significantly strengthen the results. Keywords: Consensus, Byzantine agreement, Clock synchronization, Distributed computing and fault tolerance., Prepared in cooperation with Massachusetts Inst. of Tech., Cambridge, Contracts N00014-85-K-0168, N0014-83-K-0125 , DAAG29-84-K-0058 and Grant NSF-DCR83-02391.