Paper #5.2.20

Indexing techniques for data broadcast on wireless channels

Problems

Power conservation and access efficiency are two of the most important
characteristics of a wireless data dissemination service. Data indexing
techniques have show to reduce power consumption significantly at the 
cost of some access time overhead. The paper introduces two access efficient 
indexing methods called cache schedule and integrated signature for a wireless system
that offers both broadcast and on-demand services.

Strengths

(1) The paper describes the different kinds of services (on-demand, broadcast) offered
by mobile data-dissemination networks. It also discusses the advantages of one over
the other such as power usage and data access efficiency of each type of service.

(2) Since the two access modes (on-demand & broadcast) have different advantages based
on the network load, wireless data channels can be programmed to be assigned to one or
to the accordingly. The paper discusses the different ways (static hybrid, dynamic hybrid, 
exclusive on-demand, exclusive broadcast) in which data channels can
be utilized so that data channel utilization is most efficient.

(3) The paper also discusses the different index-tree as well as signature indexing methods.
It describes in detail how these indexing methods work and the tradeoffs between the methods.
Tradeoff discussion includes things like signature methods is better for localized and serialized
data access whereas the index-tree method is better for a more global view but for random access
of data.

(4) The authors propose a hierarchical view for mobile data access that combines the different
indexing techniques for overall improvement in data access time. The idea is that the most commonly
accessed data items can be cached, the commonly interesting data subset can reside on the broadcast
channel and the rest of the data can be stored on the server and offered only through the on-demand
mode. This is a great concept that in theory offers the best utilization of the data channel.

Weaknesses

(1) There is no actual implementation of such a data delivery system. While the theoretical ideas are
strong, there is no way of validating the claims.

(2) The hybrid system is heavily dependent on the client data access patterns as the system is 
supposed to adapt to different loads. There should have been some discussion of how data access
patterns would be monitored and what would be the tradeoff and costs involved in using such
a system.