Keeping Track of the Latest Gossip in Message-Passing Systems
Consider a distributed system in which processes exchange information by passing messages. The gossip problem is the following: Whenever a process q receives a message from another process p, q must be able to decide which of p and q has more recent information about r, for every other process r in the system. With this data, q is in a position to update its knowledge about the global state of the system.
We propose a solution where each message between processes carries information about the current state of knowledge of the sender. This information is uniformly bounded if we make reasonable assumptions about the number of undelivered messages present at any time in the system. This means that the overhead of maintaining the latest gossip is a constant, independent of the length of the underlying computation.
KeywordsTransitive Closure Primary Graph Late Information Primary Information Message Complexity
Unable to display preview. Download preview PDF.
- D. Dolev, N. Shavit: Bounded concurrent time-stamps are constructible, Proc. ACM STOC (1989) 454–466.Google Scholar
- A. Israeli, M. Li: Bounded time-stamps, Proc. 28th IEEE FOCS (1987) 371–382.Google Scholar
- R. Krishnan, S. Venkatesh: Optimizing the gossip automaton, Report TCS- 94–3, School of Mathematics, SPIC Science Foundation, Madras, India (1994).Google Scholar
- L. Lamport, N. Lynch: Distributed Computing: Models and Methods, in: J. van Leeuwen (ed.), Handbook of Theoretical Computer Science: Volume B, North-Holland, Amsterdam (1990) 1157–1200.Google Scholar
- M. Mukund, K. Narayan Kumar, M. Sohoni: Keeping track of the latest gossip in message-passing systems, Report TCS-95-3, School of Mathematics, SPIC Science Foundation, Madras, India (1995).Google Scholar