Information Technology Reference
In-Depth Information
Extended Retrieval Protocol
1) Probe the channel and retrieve the offset to the next index.
2) Access the next index
3) Do {Search the index for the requested object
4) Calculate the offset of the object
5) Get the channel on which the object will be broadcast
6) } while there is an unprocessed requested object
7) Generate access patterns for the requested objects (using retrieval scheme)
8) Do {Wait for the next broadcast cycle
9) Do {Reach the first object as indicated by the access pattern
10) Retrieve the object
11) } while there is an un-retrieved object in the access pattern
12) } while there is an unprocessed access pattern
7.2.1 Performance Evaluation
To validate the feasibility of this approach, to compare and contrast them against
each other, and to show the trade-off between access latency and computational com-
plexity the scope of the simulator reported in Section 6 was extended. The extended
simulator emulates the process of accessing data from a hierarchical indexing scheme
for both the bottom up and the top down approaches discussed in Sections 3.6.1 and
3.6.2 , respectively, in addition, it analyzes the effect of conflicts on the average ac-
cess time and power consumption.
The index structure can be transmitted in different fashions including:
A complete index is transmitted at the beginning of each broadcast in the first
channel before the data;
Index is distributed among the data elements;
Index is replicated and interleaved with the data elements; or
A dedicated channel is used to exclusively and continuously transmit the index
in a cyclic manner.
The experimental results indicated that repeated transmission of the index on a
separate channel provides the best response time. Hence, the following discussion is
limited to the employment of dedicated channel to broadcast the index structure.
User requests were randomly generated representing a collection of K objects in
the broadcast. In various simulations runs, the value of K was varied from one to
N × M —in a typical user query of public data, K is much less than N × M . Finally,
to take future technological advances into account, parameters such as transmission
rate and power consumption in different modes of operation were fed to the simulator
as variable entities.
 
Search WWH ::




Custom Search