A Gentle Wakeup Call: Symmetry Breaking with Less Collision Cost

Abstract: The wakeup problem addresses the fundamental challenge of symmetry breaking. There are $n$ devices sharing a time-slotted multiple access channel. In any fixed slot, if a single device sends a packet, it succeeds; however, if two or more devices send, then there is a collision and none of the corresponding packets succeed. For the static version of wakeup, all packets are initially active (i.e., can send and listen on the channel); for the dynamic version, the packets become active at arbitrary times. In both versions, the goal is to successfully send a single packet. Prior results on wakeup have largely focused on the number of slots until the first success; that is, the latency. However, in many modern systems, collisions introduce significant delay, an aspect that current wakeup algorithms do not address. For instance, while existing results for static wakeup have polylogarithmic-in-$n$ latency, they can incur additional latency that is {\it linear} in the cost of a collision $C$. Thus, the total latency is large and dominated by the contributions from collisions. Here, we design and analyze a randomized wakeup algorithm, Aim-High. For sufficiently large $C$ and with bounded error, Aim-High has latency and expected collision cost that is nearly $O(\sqrt{C})$ for both the static and dynamic versions. Otherwise, the latency and expected collision cost are $O(\texttt{poly}{(\log n)})$ for the static setting, and $O(n\,\texttt{poly}{(\log n)})$ for the dynamic setting. We also establish lower bounds that complement these results.