A clock synchronization problem with random delays (Q1121019): Difference between revisions
From MaRDI portal
Changed an Item |
Set OpenAlex properties. |
||
(2 intermediate revisions by 2 users not shown) | |||
Property / MaRDI profile type | |||
Property / MaRDI profile type: MaRDI publication profile / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Q5532825 / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Optimal precision in the presence of uncertainty / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: An upper and lower bound for clock synchronization / rank | |||
Normal rank | |||
Property / full work available at URL | |||
Property / full work available at URL: https://doi.org/10.1016/0885-064x(89)90009-5 / rank | |||
Normal rank | |||
Property / OpenAlex ID | |||
Property / OpenAlex ID: W1979607558 / rank | |||
Normal rank |
Latest revision as of 10:43, 30 July 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | A clock synchronization problem with random delays |
scientific article |
Statements
A clock synchronization problem with random delays (English)
0 references
1989
0 references
The clock synchronization problem is the problem to find an algorithm which shifts the clocks so that the error, i.e., the largest difference between any two shifted clocks, is possibly small. To synchronize the clocks, processors communicate by sending local clocks readings. The time to transmit a message is considered to be a random variable with a known probability distribution. It is assumed that the delays are identically distributed on each communication link, and that the clocks do not drift. The following questions are examined: (1) For a network with k processors and a fixed number n, what is the minimal average error among all algorithms that use n messages? (2) What is the algorithm whose average error is minimal? For the case of two processors the n-th minimal error is derived and the optimal algorithm is described. For the general case of k processors, lower and upper bounds are determined for the n-th minimal error, and a near optimal algorithm is described. The algorithms are free of deadlocks, robust in a specific sense, and can be implemented so that all messages except k-1 are single bit messages. The algorithms are fully parallelizable.
0 references
random delays
0 references
distributed networks
0 references
parallelization
0 references
clock synchronization
0 references