Database Reference
In-Depth Information
between the different records of the fact table. The map table links
together the different records in the table which form a path.
We note that the map table is the only additional information which
needs to be maintained in the case of an RFID data warehouse. The
RFID warehouse can be viewed as a multi-level database, in which the
lowest level of representation are the raw RFID records, whereas the
higher levels contain the cleansed and compressed records. In addition
to the use of group-wise movements for compressing the data, a variety
of other abstractions can be used for further compression. For exam-
ple, if a minimum time granularity of one hour is required, then the
set of movements and stays occurring in a single hour can be consoli-
dated into a single movement. Similarly, the location can be specified
at a higher level of granularity, and the sizes and the types of products
can also be consolidated. This is because users are often interested in
queries at much higher abstraction levels. Many path segments which
are less important can also be eliminated and consolidated into a single
movement. For example, for a store manager, the movement of items
between shelves may not be important, and can either be eliminated or
consolidated with some other movement. All of these operations signifi-
cantly reduce the size of the representation, and make higher level query
processing operations much more ecient.
Some other characteristics of the different kinds of tables such as the
info table
,
stay table
,and
map
table are as follows:
The
info
table contains path-independent dimensions. Each di-
mension can have an associated concept hierarchy on which OLAP
operations can be performed. For example, one could drill down
on a particular product category and support aggregate queries on
this category.
The
stay
table contains the
TimeIn
and
TimeOut
information for
the different products. In order to save space, this information is
stored in terms of aggregated GIDs of items which move together,
rather than the individual EPC values.
The map table contains the hierarchy of GIDs in the data. Each
entry is of the form (
gid,
(
gid
1
...gid
r
)). This implies that
gid
points to
gid
1
...gid
r
.Wenotethatatthelowerlevels,
gid
k
could
correspond to an individual EPC. The higher levels of the gid, are
also labeled with locations, with one identifier corresponding to
each location for items in the gid.
We note that the use of the gids, as maintained by the mapping ta-
ble can provide a very ecient way to perform the queries, since each
