Abstract
Recently, the demand for real-time data services has been increasing (Aslinger & Son, 2005). Many applications such as online stock trading, agile manufacturing, traffic control, target tracking, network management, and so forth, require the support of a distributed real-time database system (DRTDBS). Typically, these applications need predictable response time, and they often have to process various kinds of queries in a timely fashion. A DRTDBS is defined as a distributed database system within which transactions and data have timing characteristics or explicit timing constraints and system correctness that depend not only on the logic results but also on the time at which the logic results are produced. Similar to conventional real-time systems, transactions in DRTDBSs are usually associated with timing constraints. On the other hand, a DRTDBS must maintain databases for useful information, support the manipulation of the databases, and process transactions. Timing constraints of transactions in a DRTDBS are typically specified in the form of deadlines that require a transaction to be completed by a specified time. For soft realtime transactions, failure to meet a deadline can cause the results to lose their value, and for firm or hard real-time transactions, a result produced too late may be useless or harmful. DRTDBSs often process both temporal data that lose validity after their period of validity and persistent data that remain valid regardless of time. In order to meet the timing constraints of transactions and data, DRTDBSs usually adopt main memory database (MMDB) as their ground support. In an MMDB, “working copy” of a database is placed in the main memory, and a “secondary copy” of the database on disks serves as backup. Data I/O can be eliminated during a transaction execution by adopting an MMDB so that a substantial performance improvement can be achieved. We define a DRTDBS integrating MMDB as a distributed real-time main memory database system (DRTMMDBS).
TopIntroduction
Recently, the demand for real-time data services has been increasing (Aslinger & Son, 2005). Many applications such as online stock trading, agile manufacturing, traffic control, target tracking, network management, and so forth, require the support of a distributed real-time database system (DRTDBS). Typically, these applications need predictable response time, and they often have to process various kinds of queries in a timely fashion. A DRTDBS is defined as a distributed database system within which transactions and data have timing characteristics or explicit timing constraints and system correctness that depend not only on the logic results but also on the time at which the logic results are produced. Similar to conventional real-time systems, transactions in DRTDBSs are usually associated with timing constraints. On the other hand, a DRTDBS must maintain databases for useful information, support the manipulation of the databases, and process transactions. Timing constraints of transactions in a DRTDBS are typically specified in the form of deadlines that require a transaction to be completed by a specified time. For soft real-time transactions, failure to meet a deadline can cause the results to lose their value, and for firm or hard real-time transactions, a result produced too late may be useless or harmful. DRTDBSs often process both temporal data that lose validity after their period of validity and persistent data that remain valid regardless of time. In order to meet the timing constraints of transactions and data, DRTDBSs usually adopt main memory database (MMDB) as their ground support. In an MMDB, “working copy” of a database is placed in the main memory, and a “secondary copy” of the database on disks serves as backup. Data I/O can be eliminated during a transaction execution by adopting an MMDB so that a substantial performance improvement can be achieved. We define a DRTDBS integrating MMDB as a distributed real-time main memory database system (DRTMMDBS).
The existing researches on DRTDBS focus mainly on concurrency control (Gustafsson, Hallqvist & Hansson, 2005; Lam, Kuo, Tsang & Law, 2000; Ulusoy, 1993), replication (Aslinger & Son, 2005; Son & Kouloumbis, 1993; Ulusoy, 1994), and commitment (Haritsa, Ramamritham & Gupta, 2000; Qin & Liu, 2003; Xiao, Liu, Deng & Liao, 2006). The studies of failure recovery for a DRTMMDBS are relatively scarce. However, due to the complexity of distributed environments together with volatility and vulnerability of the main memory, the possibility of failure in a DRTMMDBS becomes much larger than in centralized disk resident database systems. When a site crash occurs, as the databases are not available for transactions, many transactions may miss deadlines, and a large amount of temporal data may lose their validity before they can be used, so a DRTMMDBS should have the ability of high fault-tolerance and can resume services again as quickly as possible after crashes.
Key Terms in this Chapter
Sequential Logging: A traditional logging technique. The logging technique requires storing log records into a single logging file in sequence according to the execution order of concurrent transaction operations.
Deadline: In a real-time database system or a distributed real-time database system, deadline is used to describe the timing constraint of transaction. The deadline of a transaction defines the specified time by which the transaction must be completed.
Temporal Data Object: Used to record the status of external objects in the world. Each temporal data object is associated with a period of validity. A temporal data object will lose validity after its period of validity.
Distributed Real-Time Main Memory Database System (DRTMMDBS): The distributed real-time database system that adopts a main memory database as its ground support.
Distributed Real-Time Database System (DRTDBS): A distributed database system within which transactions and data have timing characteristics or explicit timing constraints, and system correctness depends not only on the logic results but also on the time at which the logic results are produced.
Main Memory Database (MMDB): The database where the primary copy of a database is placed in main memory and a “secondary copy” of a database on disks serves as a backup.
System Failure: Also called crash, occurs when the contents of volatile main memory storage are lost or corrupted.
Partitioned Logging: A novel logging technique relative to the sequential logging. Partitioned logging stores log records according to transaction class (data class); that is, log records belonging to different transaction class (data class) are stored in different partitions.