This paper proposes two improvement to the Calendar Queue. The improvement is targeting at the resizing mechanism.
First improvement is to detect uneven event distribution. The conventional resize rule checks the total events in a queue and resize if it outsides a bound. This suggested improvement keeps encountered event count which sums the number of events seen in those buckets with new event enqueued. This may double count a bucket but the objective is to calculate the average encountered events which can infer uneven event distribution. If that average is above a threshold, resize the calendar queue with bucket size changed.
Second improvement is to find the best bucket size. It keeps track on the bucket with longest event queue and the event queue length in each bucket. Then the buckets around the longest bucket are considered to compute the average inter-event gap and this is the new bucket width.
Bibliographic data
@inproceedings{
title = "Dynamic Calendar Queue",
author = "SeungHyun Oh and JongSuk Ahn",
booktitle = "Proceedings of the 32nd Annual Simulation Symposium",
pages = "20--25",
month = "Apr",
year = "1999",
}