Bernhard Haeupler and Dahlia Malkhi
Gossip algorithms spread information in distributed networks by having nodes repeatedly forward information to a few random contacts. By their very nature, gossip algorithms tend to be distributed and fault tolerant. If done right, they can also be fast and message-efficient. A common model for gossip communication is the random phone call model, in which in each synchronous round each node can PUSH or PULL information to or from a random other node. For example, Karp et al. [FOCS 2000] gave algorithms in this model that spread a message to all nodes in Θ(log n) rounds while sending only O(log log n) messages per node on average. They also showed that at least Θ(log n) rounds are necessary in this model and that algorithms achieving this round-complexity need to send ω(1) messages per node on average.
Recently, Avin and Elsässer [DISC 2013], studied the random phone call model with the natural and commonly used assumption of direct addressing. Direct addressing allows nodes to directly contact nodes whose ID (e.g., IP address) was learned before. They show that in this setting, one can “break the log n barrier” and achieve a gossip algorithm running in O(√log n) rounds, albeit while using O(√log n) messages per node.
In this paper we study the same model and give a simple gossip algorithm which spreads a message in only O(log log n) rounds. We furthermore prove a matching Ω(log log n) lower bound which shows that this running time is best possible. In particular we show that any gossip algorithm takes with high probability at least 0.99 log log n rounds to terminate. Lastly, our algorithm can be tweaked to send only O(1) messages per node on average with only O(log n) bits per message. Our algorithm therefore simultaneously achieves the optimal round-, message-, and bit-complexity for this setting. As all prior gossip algorithms, our algorithm is also robust against failures. In particular, if in the beginning an oblivious adversary fails any F nodes our algorithm still, with high probability, informs all but o(F) surviving nodes.
|Published in||PODC 2014|