Database Development

The core of NDBHMI consists of a distributed database system storing the available hydrology - related data of the country. Each data provider (namely public sectors) has its own local database in his premises. The different databases are linked to each other through network links. This network / database topology design proved to be the appropriate one as far as uninterrupted operability of the system and data security and isolation is concerned. The main aspects of the database design were the following:

Data reliability, integrity, availability

Schematic representation of data check procedure in non-variable data Data conversion (from their original format in data forms and autographic instrument tapes to digital files) proved to be the most error prone procedure in the implementation of the database. Special checks were developed and applied to the data by the project team to face the problem of erroneous migration of data including checks of extreme values, spatial and time domain consistency etc. Different checks were applied to constant data such as station names, coordinates, gauging instruments etc. in order to ensure both their correctness and their consistency with the associated timeseries data. Data integrity in the database level is enforced through the use of foreign keys.

Data security - Enforcement of charging rules

NDBHMI WAN (Wide Area Network) One of the business rules that the database design had to obey to was the data ownership. Each data provider must have the responsibility as well as the flexibility of managing their own data. This business rule led to the implementation of a distributed database system in which data are stored localy in the data providers databases and these databases communicate with each other through database links using the existing Wide Area Network. This network topology is transparent to the end user of the system who - provided that he is connected to any node of the system - can query and retrieve data. However the retrieval of the data is governed by rules imposed at the local database level, i.e. PPC may reject requests for data retrieval from a given user, whereas NMS allows this user to access its data. This way, each provider can independently apply his own charging and security policies to its data without affecting the other nodes (and data stored there) of the system. At the local database level, data security is enforced through the use of database roles.

Database performance

Timeseries data may extend to more than 100-year continuous recordings. For a recording interval of a day (that is a daily timestep) the whole timeseries consists of more than 364 * 100 = 36400 recordings and if one takes into account the fact that the use of 20 - 30 timeseries simultaneously is common in hydrology (estimation of mean areal precipitation, thiessen polygons etc.), the need for maximum database performance seems obvious. These considerations were taken into account during the database development phase leading to load balancing schemes and per session optimizations.

Storage considerations

The great amount of information that had to be stored in the database led to some additional considerations as far as storage was concerned. A balanced approach between performance issues and storage requirements was followed in order to reach the optimum relation. In fact the chosen solution proved to be both a high performance one and the one that minimises storage requirements. Crucial database tables were built as "index-organised" ones. This feature of the database is used when the storage must be minimised without compromising query and retrieval performance.

Friendly user interface

Database management application (DBManager) The DBManager (Database manager) application is the interface of the database to the user. The program exploits the database design and all the specific features of the NDBHMI database in order to give the user the flexibility of a tool specifically designed for this purpose. Data can be queried using both hydrological and conventional criteria and retrieved obeying to the specific policy of each public sector that owns the data. A thin client - fat server model has been followed in order to provide a more flexible environment when it comes to new additions and changes in the underlying policies.

Extensibility - Links to other relevant data sources

The linking of the NDBHMI database to other relevant databases and data sources was considered a very important aspect from the early stages of the design phase of the project. As a consequence an abstract interface has been built in the database level and currently a link to the National Network of Environmental Informatics database is on line giving users of one of the two systems the ability to navigate both databases.