We study the wakeup problem in an asynchronous system of anonymous processes that are identically coded. Informally, a protocol solves wakeup if, in all infinite fair runs of the protocol, every process eventually learns that every process has executed at least one step and is therefore ‘awake’[FMRT90]. We study this problem in the context of two models of shared memory communication. In the first model, each word of shared memory may be read and written by every process. In the second model, each process has a register which only that process may write, but all processes may read. We show that wakeup is not solvable in the first model. For the second model, we derive matching lower and upper bounds on the amount of shared memory necessary to solve wakeup. We also study the feasibility of leader election and consensus in the above two models.
[1]
Edsger W. Dijkstra,et al.
Self-stabilizing systems in spite of distributed control
,
1974,
CACM.
[2]
Michael J. Fischer,et al.
The wakeup problem
,
1990,
STOC '90.
[3]
Nancy A. Lynch,et al.
Impossibility of distributed consensus with one faulty process
,
1985,
JACM.
[4]
Amos Israeli,et al.
On processor coordination using asynchronous hardware
,
1987,
PODC '87.
[5]
Nancy A. Lynch,et al.
Impossibility of distributed consensus with one faulty process
,
1983,
PODS '83.
[6]
Alon Itai.
On the computational power needed to elect a leader
,
1990
.