The two types of HiRDB are the HiRDB/Single Server that operates on a single server machine, and the HiRDB/Parallel Server that operates on two or more server machines. You can select either a HiRDB/Single Server or a HiRDB Parallel Server, depending on the scope of the operations to which HiRDB will be applied. You can change a HiRDB/Single Server into a HiRDB/Parallel Server, or you can increase the number of server machines used in a HiRDB/Parallel Server. Thus, you can expand your system gradually as the scale of your operations increases.

The Shared Nothing method that is incorporated into HiRDB is well-suited for parallel processing and enables you to increase the processing capacity of HiRDB in proportion to the number of processors that are used. In other words, you can increase processing capacity without having to add server machines. When the flexible hash partitioning function is used and a server machine is added, HiRDB modifies the hash function automatically, so that the data is stored automatically in the new server machine.

(a) Performance improvement through parallel processing

Concurrent retrieval and updating of tables

Database search and update operations can be distributed to multiple server machines, and table data can also be divided among the server machines. This feature permits concurrent searching and updating of tables, which improves performance in direct proportion to the number of server machines deployed.

Distribution of high-overhead database processing operations

HiRDB enables you to assign high-overhead operations, such as sorting and joining, to a separate server machine so that these operations can be performed concurrently with data accesses. This feature reduces the amount of time required to output search results.

Reduction in batch processing time for large amounts of data

HiRDB can reduce the amount of time required for processing because it can store masses of data concurrently, such as during system creation.

Concurrent database reorganization

HiRDB can reduce the amount of time required for database reorganization operations because it can concurrently reorganize databases by server.

Concurrent backup and recovery operations

In HiRDB, a single command can concurrently back up and recover a failed database. This feature reduces the amount of time required for backup and database recovery operations.

(b) Fine buffer control through the use of global buffers

HiRDB enables you to allocate index data and other frequently accessed data to special buffers. This feature assures stable response by eliminating interference between buffers even in an environment where different operations are performed on a mixed basis, such as index searches and complete data searches.

(c) Performance improvement by using synchronization point dump processing

In an operation called synchronization point dump processing, HiRDB stores update information up to a specific point in the database and recovery information up to that point in a file. Other systems stop accepting transactions during a synchronization point dump operation, which results in a decrease in processing performance. HiRDB, by contrast, does not limit the system's ability to accept transactions, so there is no decrease in processing performance attributable to synchronization point dump processing.

(d) Rapid system recovery by high-speed rerun

HiRDB reduces the range over which a recovery operation must be performed in the aftermath of a system failure by periodically collecting synchronization point dumps, which enables HiRDB to quickly complete recovery processing.

In addition, HiRDB performs during a system recovery operation a delayed rerun operation that simultaneously starts a rollback operation and acceptance of new transactions, which can speed up system restart.

(a) Storing in files information needed for system restart and for database recovery

Logging information needing for system restart

During operation, HiRDB stores in files system status information that will be needed to restart the system in the event of a failure. These files are called status files.

Logging information needed for database recovery

HiRDB stores in files a history of database update information (a system log) that would be needed for recovery of the database during a recovery operation. The files in which this information is stored are called system log files. In the event of an error in the system, system log files make it possible for HiRDB to correctly recover the database to its status just before the error occurred.

Duplicate files

HiRDB maintains duplicate status files and system log files, thus providing redundant logging of the information it will need for system and database recovery. File redundancy permits the use of the other file if a problem arises in one of the files, thus increasing system reliability.

(b) Automatic system restart

In the event of a relatively minor error, HiRDB uses the status files to restart the system automatically, so there is no need for operator intervention.

(c) Reducing system downtime by use of system switchover facility

In HiRDB, you can set up a standby server machine that is separate from the currently operating server machine, so that if the current server machine fails, operations can be switched smoothly to the standby machine. This feature is called the system switchover facility.

Microsoft Cluster Server (MSCS) is used for the system switchover facility.

(a) Intensive control from a specific server machine

In the case of a HiRDB/Parallel Server, a specified server machine can integrally and intensively control a HiRDB system running on multiple server machines. For example, you can execute a command or a utility on one server machine to start or terminate the HiRDB running on all server machines or on specified server machines.

(b) Support for setting up the HiRDB environment

Support tools are provided for setting up the HiRDB environment. The tools that are provided are listed and described below.

Simple setup tool

A graphic user interface (GUI) is provided for setting up HiRDB. Choosing Standard Setup, in which you simply specify the setup directory, enables you to easily configure an environment. You can also choose Custom Setup in order to specify more detailed settings. The simple setup tool can also be used to update or edit system definitions that you have already created.

Batch files

By executing the batch files, you can automatically set up the basic HiRDB environment.

(a) XA interface for X/Open

HiRDB can communicate with OLTP by means of the XA interface of X/Open. A HiRDB XA library is provided so that HiRDB transactions can be controlled by the Transaction Manager.

(b) ODBC, JDBC, and OLE DB interfaces

HiRDB complies with the ODBC, JDBC, and OLE DB industry standards, which means that ODBC, JDBC, and OLE DB applications can be used in HiRDB. You can also use ADO (ADO.NET is also supported), DAO, and RDO.

With the growing popularity of Internet businesses, there is an increasing need to conduct online operations non-stop, 24 hours per day, 365 days per year. HiRDB provides functionality based on the premise of continuous operation, 24 hours per day. For details about functionality based on 24-hour-per-day operation, see the HiRDB Version 8 System Operation Guide.