EXtremeDB

From LQWiki
Jump to: navigation, search

eXtremeDB, developed by McObject,is a high performance, low-latency, ACID-compliant embedded database management system using an in-memory database system (IMDS) architecture and designed to be linked into C/C++ based programs. It works on Windows, Linux, and other real-time and embedded operating systems. Building on this core IMDS, there are several editions supporting high availability, database clustering, hybrid (in-memory and disk-based) data storage, columnar layout of time series data (such as financial market data) and other specialized abilities.

Product features

In-memory data storage

eXtremeDB provides direct access to data, which is stored entirely in memory. This eliminates disk reads and writes from the data management process, and removes the need for complex software logic such as database caching functions. With this simplified design, eXtremeDB provides a code “footprint” as small as 50K, and typical database read and write accesses at the level of a few microseconds, or less.

Transactions

For data integrity, eXtremeDB transactions support the ACID properties, ensuring that operations grouped into transactions will complete together or the database will be rolled back to a pre-transaction state.

APIs

Native API

eXtremeDB's native application programming interface (API) includes a standard function library for basic database operations, as well as data manipulation functions that are generated when the database scheme is compiled, and which derives from the application’s data model. Database functions therefore reflect the purpose and schema for which the database is being used. The native API is type-safe – the developer's C/C++ compiler will catch data typing and assignment errors when the application is built.

SQL

eXtremeDB offers a SQL API, called eXtremeSQL, which implements much of the ANSI SQL-89 specification. eXtremeSQL includes eXtremeDB-specific extensions including support for structure], arrays and vectors, as well as query optimizations based on specific eXtremeDB capabilities.

XML

eXtremeDB provides XML Extensions to facilitate simple schema evolution and the exchange of data between the eXtremeDB embedded database and external systems.

Developer tools

Typical eXtremeDB users are software developers creating applications in the C and C++ languages, who embed eXtremeDB’s data management functions in their application code, with the goal of maximum run-time efficiency. eXtremeDB features for coding in these languages include:

  • Support for virtually all data types, including structures, arrays, vectors and binary large objects (BLOBs)
  • Querying methods include hash indexes for exact match searches
  • Tree indexes that support queries for pattern match, range retrieval and sorting
  • “Voluntary” indexes for program control over index population
  • Object identifier references provide direct data access
  • Autoid for system-defined object identifiers
  • Indexes that contain only a reference to data, rather than storing duplicate data, in order to minimize memory requirements
  • Synchronous/asynchronous event notifications
  • Object history

High availability

For software applications that require high availability, eXtremeDB High Availability (HA) Edition is designed to ensure that changes to a master database and identical standby databases succeed or fail together, and enables deployment of multiple fully synchronized eXtremeDB-HA databases within the same hardware device or across multiple, widely distributed systems. A predictable response time is created via the HA protocol's “time cognizance” – if a master or replica database fails to respond within a pre-set limit, it is automatically decommissioned and replaced in its role by a responding eXtremeDB-HA database instance.

Transaction logging

The eXtremeDB Transaction Logging Edition provides additional recovery capabilities. When transaction lo]ging is active and enabled, all updates to data objects are logged to a set of files on disk, or on a network device. If the memory content is damaged or destroyed, the exact state of the database is restored by an automatic roll forward procedure.

64-Bit database

eXtremeDB-64 increases maximum database size by adding support for 64-bit micro-processors and operating systems. Compared to traditional 32-bit processors, 64-bit technology approximately doubles the amount of data a CPU can handle per clock cycle, and increases the amount of memory a system can address from approximately 3GB to more than one terabyte (1000GB).

Hybrid storage (eXtremeDB Fusion)

eXtremeDB Fusion is marketed by McObject as a "hybrid" embedded database system, in that it provides both in-memory and on-disk data storage in a single database instance. A notation in the database schem] causes specified records to be managed entirely in main memory, while others are saved to disk. eXtremeDB Fusion's on-disk data management process uses database caching to keep frequently requested records in memory (while flushing updates through to permanent storage), while records specified as in-memory are accessed directly, eliminating mechanical disk I/O, multiple data copies, and (now redundant) caching processes.


Benefits of this hybrid approach cited by McObject include performance gains when sorting, storing and retrieving specified data entirely in memory, rather than going to disk; cost, because a less expensive hard disk can be substituted for more expensive RAM as a data storage medium; persistence; and form factor, because RAM chips can’t yet approach the density of an 80GB micro-drive.

Kernel mode

In 2008, McObject released eXtremeDB Kernel Mode (KM), an embedded database system designed for deployment in the operating system kernel. The company described eXtremeDB-KM's intended use as providing local data management for application logic that is deployed in kernel mode, in order to eliminating the context switches (and their performance penalty) that occur when a kernel mode process interacts with user mode software (in this case, a database system running in user mode). In a Dr. Dobb's Journal feature article about the new database technology, McObject CTO Andrei Gorine named security applications (such as access control systems and firewalls) and operating-system monitors as applications that commonly place functions in the operating-system kernel and would benefit from a kernel mode database.

External links