Vous êtes sur la page 1sur 2

1.2.

3
Query processing
To guarantee the data independence property, the architecture of a DBMS
is typically divided in three layers: conceptual, logical, and physical. Each
query is formulated in some calculus like language at the conceptual level.
In the commonly used relational DBMSs this language is usually SQL. Next
this high-level query expression is translated into a logical expression, which
usually is some kind of relational algebra. Finally, this algebra expression is
translated into a physical expression.
1.2.4
Query optimization
Recall our two options to evaluate the query in Section 1.2.2. Every database
system has a component called query optimizer that is responsible for taking
the decisions which way to process a query.
One can imagine that the number of employees and departments in total and
the number of both that match the query are important factors in deciding
which of both processing alternatives is the most efficient. The query optimizer
therefore is supported by a cost model. This model is able to estimate proper-
ties like cardinalities, i.e., the number of tuples in a table, and tries to estimate
how much effort it will take the system to process a given evaluation variant as
suggested by the query optimizer. To be able to properly determine the car-
dinalities of intermediate results, the cost model needs a notion of how many
tuples are selected by the operator that produces that particular intermediate
result. This notion of how much an operator selects is called selectivity. Often
this selectivity can not be computed exactly, it can only be guessed. It is the
job of the selectivity model to do this as accurately as possible.
The optimizer operates on all three architecture levels. On the conceptual level
its task is typically limited to assisting in translating the query expression into
some standardized form. At the logical level the typical operator reordering
problems have to be solved, like the one in our salary example. When the
logical expression is translated into the physical form the optimizer for example
has to take into account whether index structures, i.e., special data structures
designed to speed-up access to certain parts of the data, are present or not, or
whether the data is stored in some ordered format.
1.3
Information retrieval
Similar to our introduction to the database field in the previous section this
section provides a very basic introduction to some key principles in the infor-
mation retrieval field. An experienced information retrieval researcher might
choose to skip this section and go directly to the next one.
This section is structured as follows. First, we sketch the typical application
domain where information retrieval technology is being used, in Section 1.3.1.
Next, in Section 1.3.2, we introduce some basics concerning the ranking of
documents.

Vous aimerez peut-être aussi