Research on database system under grid environment

The combination of database technology and other computer technology can produce new research fields. For example, the combination of database technology and object-oriented technology results in object-oriented database, and the combination of database technology and grid technology results in a new research content called Grid database. The current main research contents of grid database include three aspects: Grid database management system, grid database integration and support new grid application.

Grid database management System

The grid provides a platform that supports systematic identification and authorization, resource discovery, data transfer, process creation and scheduling, and dynamic binding across heterogeneous platforms. On the basis of these functions provided by grid, a high performance grid database management system with safe and reliable capability and independent computing ability can be constructed. Grid database management system will eventually become an important resource on the grid, providing data management services.

In the construction of grid database management system, the first step is to provide a middleware, the database management system packaging into grid services, so that grid applications access to the grid database. The advantage of this approach is that there is basically no need to make any changes to the existing database management system.

Further work is to expand the existing database management system, so that it can directly use the functions provided by the grid to achieve the distributed database and related grid services. Here are some things you can do to extend an existing database management system:

On the grid, users can use all the available resources on the grid by simply logging in once. The grid database management system needs to support the identity authentication and authorization mechanism of this consistency in grid.

Many grid applications need to receive a notification of a task completion after submitting a task to the database management system, which needs to combine the trigger mechanism of the grid database management system and the notification service in the grid.

In a grid, query results can be cached for transmission to third parties, which means that, within an indeterminate period of time, the grid database management system must retain the results and then transmit the query results to a third party using the grid data transmission mechanism.

In the grid, distributed query processing can use the grid resource discovery mechanism to understand the status of the network, get the right resources to run distributed queries or store intermediate results.

The grid database management system should provide resource reservation function to participate in the process scheduling in the grid.

The grid database management system should be able to provide the necessary information for the grid accounting and payment standards so that users can pay for the use of resources.

Some grid problems are not solved by simply expanding existing database management systems. Here are a few examples to illustrate this point:

The coupling between computing and data resources is very close to the grid, which is a challenge for database query optimization. Computing tasks where to execute, need not to migrate, how data is stored, where to store, how to dispatch the data on the operation and calculation tasks, short-term optimization or long-term optimization, etc., all need to be in the query optimization to give a holistic, multi-angle considerations.

In a grid, when the size of the data grows to a certain extent, the computational code transmitted on the data is more economical than the transmission of all the data. The calculations on the data are more complex than the database-supported data queries, which generate a large number of intermediate results, requiring a large number of processors and transferring large amounts of data between multiple grid nodes. The calculations on the data also need to be monitored. The current database management system is not ready to host this large calculation, nor does it support the required monitoring interfaces.

The traditional database management system is a complex system, and the runtime consumes a lot of resources as a whole. A grid database management system is best composed of components that can be combined to perform some or all of the functions of the database management system as needed. In addition to reducing resource consumption, it is more important to make it possible to optimize the use of database resources based on the overall system size.

Use the registered method in the grid to publish and find resources. There is a lot of metadata stored in the database, it is not feasible to register so many metadata manually, but how to complete autoenrollment? The problem with autoenrollment is not to register the metadata in the grid, but to ensure that the metadata is understood by the third party when registering. This problem is also a problem that needs to be solved when the grid database is integrated.

Grid Database Integration

Because the main goal of Grid is to support collaborative work on shared resources, grid database integration is the hotspot of grid data management at present. Grid database integration is the use of information from two or more grid databases and the use of this information to build a large database.