Information Technology Reference
In-Depth Information
and by assumption (6)
L
=1
b
B
.
(6)
u i
k
i
The class division is not taken into consideration. The second phase is similar to the
third phase presented in the class-based auction approach.
Some algorithms were developed to increase the income of web service and for the
client satisfaction, and were based on the second-highest price (Vickrey) approach.
5 Results of Simulation
Three algorithms: Class-based Auction Algorithm, Profitability based Auction Algo-
rithm and FIFO Scheduling Algorithm were implemented in the Matlab environment.
The stream of offers was simulated as a Poisson Process. For different number of of-
fers, and different available bandwidths (10Mbps and 62.5 Mbps) the profit of web
service was measured and compared.
Fig. 1 and Fig. 2 present the performance index defined as the sum of the income
of web service vs. number of serviced offers. The biggest profit (service income) was
observed for Profitability based Auction Algorithm (depicted by squares) and a bit
worse for Class-based Auction Algorithm (depicted by rhombus). The results for
FIFO Algorithm (depicted by triangles) were always the worst. For the increasing
number of offers the difference between Profitability based Auction Algorithm and
Class-based Auction Algorithm was practically fixed whereas the FIFO algorithm
stands out and the difference to others is increasing for growing number of offers.
Fig. 1. Web service profit versus number of
offers for 10 Mbps available bandwidth.
Fig. 2. Web service profit versus number of
offers for 62.5 Mbps available bandwidth.
6 Conclusions
In this paper two new scheduling algorithms based on auction approach were pre-
sented. The classification of offers was depicted in an example. The profitability in-
dex was shown. The idea of scheduling of offers was shown and discussed. Some fea-
tures of presented mechanisms were distinguished.
Search WWH ::




Custom Search