Information Technology Reference
In-Depth Information
The simulator generates the average time spent on each of the three steps of the re-
trieval process (see Section 6 ). For each step, the unit switches to active mode to per-
form the action, and then returns to doze mode. Hence, the simulator calculates the
average active time when the mobile unit is in active mode accessing data and the av-
erage idle time when the unit is in doze mode. It also estimates the average query re-
sponse time. Along with the average times reported, the simulator determines the av-
erage number of broadcast passes required to retrieve objects, as well as the number
of channel switches performed by the mobile unit while retrieving data. Finally, the
simulator uses the collected information to determine the energy consumption of the
retrieval process. As a note, the size of the index was 13.52% of the size of the data
objects (not including the index) and the number of data channels varied from 1 to 16.
7.2.2 Simulation Model
For each simulation run, a request of K objects was randomly generated. A set of
input parameters including the number of parallel air channels, the broadcast trans-
mission rate, and the power consumption in different modes of operation was passed
to the simulator. The simulator was run 1000 times and the average of the designated
performance metrics was calculated.
7.2.3 Simulation Results
The retrieval protocols, discussed in Section 3.6 , are intended to reduce the num-
ber of passes over parallel channels by scheduling data retrieval. This by default
should reduce the average response time. To show this fact, in a configuration com-
posed of 2, 4, 8, and 16 channels, the bottom up retrieval scheme (see Section 3.6.1 )
and the Row Scan algorithm were simulated when the number of requested data el-
ements was varied between 5 and 50, out of 5,464 securities within the NASDAQ
exchange database, which is a reasonable range of objects requested by a query. The
simulation results showed that, regardless of the number of parallel air channels, the
bottom up retrieval scheme reduces both the number of passes and the response time
compared to the Row Scan algorithm. Moreover, the energy consumption was also
reduced, but only when the number of data elements retrieved was approximately
15 or less. For example, in an environment composed of 16 parallel air channels
when requesting 10 data elements (see Table VI ), the bottom up retrieval scheme
performed the retrieval with:
72% fewer passes,
41% reduced response time, and
3% less energy
than that of the Row Scan algorithm.
Search WWH ::




Custom Search