The Oracle TimesTen In-Memory Database (Oracle acquired TimesTen in June of 2005) is deployed in-memory in the application tier (I.E., the entire DB, including its data and its interface logic, resides in the physical memory of the machine) and as such provides enhanced performance (in comparison with disk-based DBs) to the applications that access its data. The DB supports both read and write access to in-memory data; with the vendor noting that on average the DB can respond to data read requests in 11 microseconds, and update requests in 30 microseconds. The DB can be deployed as a stand-alone offering; or--with the addition of optional software--can be used as a front-end cache to a backend Oracle 10g DB.
When deployed as a stand alone DB, the product can be accessed either directly from enabled applications (via the integration into the application of TimesTen libraries) or directly from external servers via a client/server model. The DB itself supports standard interface methods including SQL, ODBC, and JDBC. The vendor boasts a "Near Zero" administration footprint for the product, with admin tasks performed through either a programmatic interface within enabled applications, or through CLI utilities; including tools for DB backup, restore, copy, migration, and monitoring. The DB is described as both persistent and recoverable; qualities facilitated by transaction logging and database checkpoint capabilities that can be controlled by applications at the transaction level (both synchronous and asynchronous logging is supported). 数据挖掘研究院
Key optional modules for the TimesTen DB include replication (among two or more TimesTen DBs), and a Cache connect option for backend Oracle DBs. 数据挖掘研究院
The Replication option (Replication - TimesTen to TimesTen) provides a means to automatically replicate data between two or more TimesTen in-memory databases. Replication is based on a Master/Subscriber model; with bi-directional replication (servers can be both masters and subscribers for one another) and propagation (where a server subscribes to a master, and in turn operates as a master for one or more subscribers) deployments also supported. Like the DB itself, replication can be performed synchronously or asynchronously. If a DB should go down, it can automatically re-synchronize itself when it is restored. 数据挖掘研究院
The caching option--known as Cache Connect to Oracle--provides for the automated moving of data from a disk-based, backend Oracle 10g DB to an in-memory DB; allowing applications to potentially increase performance by collecting data through the cached data made available in the in-memory DB. With the cache connect option, admins can define cache groups (which consist of the target tables in the DB to be cached, and all or a subset of the rows and columns in those tables), with each cache group automatically managed by Cache Connect, or directly managed by the user application itself. Within the automated management schemes are features including support for read-only caching with automated refreshes; write-through caching (applications can both read and write to the in-memory DB) with support for asynchronous or synchronous writes; on-demand dynamic data loading where application actions can trigger the loading of specific collections of data; and automatic data aging, where data is automatically removed from the cache when it passes a certain age. Applications need connect only to the in-memory DB and are not required to connect directly to the backend DB; if the application issues an SQL request to the in-memory DB that it (the in-memory DB) cannot handle, it will pass the request through directly to the backend DB for processing.
Both options (replication and Oracle caching) are listed as fully compatible with one another. 数据挖掘研究院
New features in the latest TimesTen release include: 数据挖掘研究院
- The aforementioned on-demand data loading and automated data aging capabilities in Cache Connect 数据挖掘研究院
- Support for Oracle 10g data types 数据挖掘实验室
- Enhanced SQL functionality (the vendor notes similar semantics and behaviors as 10g)
- Support for over 50 character sets and 80 languages
The Oracle TimesTen In-Memory Database is available now, with pricing starting at $12,000 per CPU. The two options (replication and caching) start at $6,000 per CPU each. All prices depend on the size of the in-memory DB. 数据挖掘研究院
The product can be downloaded from Oracle under a developers-only licensing agreement.
