On deterministic counting in anonymous dynamic networks

Counting is a fundamental problem of every distributed system as it represents a basic building block to implement high level abstractions [2,4,6]. We focus on deterministic counting algorithms, that is we assume that no source of randomness is available to processes. We consider a dynamic system where processes do not leave the compu- tation while there is an adversary that continuously changes the communication graph connecting such processes. The adversary is only constrained to maintain at each round a connected topology, i.e. 1-interval connectivity G(1-IC) [3]. In such environment, it has been shown, [5], that counting cannot be solved without a leader. Therefore, we assume that all processes are anonymous but the distinguished leader. In the thesis we will discuss bounds and algorithms for counting in the aforementioned framework. Our bounds are obtained investigating networks where the distance between the leader and an anonymous process is persistent across rounds and is at most h, we denote such networks as G(PD)h [1]. Interestingly we will show that counting in G(PD)2 requires Ω(log |V |) rounds even when the bandwidth is unlimited. This implies that counting in networks with constant dynamic diameter requires a number of rounds that is function of the network size. We will discuss other results concerning the accuracy of counting algorithms. For the possibility side we will show an optimal counting algorithm for G(PD)h networks and a counting algorithm for G(1-IC) networks.

Nella tesi di dottorato si analizza il problema del counting in reti anonime dinamiche ed interval connesse. Vengono dimostrati lower bound non triviali sul tempo di conteggio in reti a diametro costante. Inoltre vengono sviluppati nuovi algoritmi di conteggio.