Reliable In-memory Database System

In-memory database performance

To optimize an on-disk database is to minimize disk I/O, so its design will use extra CPU cycles and extra memory to reduce or eliminate I/O.  Conversely, IMDSs by definition eliminate all disk I/O.  In other words, their optimization is primarily about delivering the highest performance at a given level of processing power.  For an in-memory database, the focus is on reducing demand for CPU cycles. Because memory is storage for an IMDS, reducing memory overhead (i.e. RAM consumed for anything other than storing data) is also a key objective.

Using eXtremeDB on embedded platforms (e.g. ARM, PowerPC), typical write accesses are at the level of a few microseconds, or less, and read-only access is sub-microsecond. When used on workstation or server platforms, performance is sub-microsecond. (Visit this page to read one of our in-memory benchmarks.)

When eXtremeDB is used as an embedded database system, the engine is reentrant, allowing for concurrent access by multiple execution tasks or threads. Used on multi-tasking platforms (e.g. Linux and VxWorks RTP), databases can be created in shared memory for concurrent access among many processes.  eXtremeDB transactions support the ACID properties, assuring data integrity.

eXtremeDB can also operate as a client/server database system. Client applications can be local (executing on the same system as the eXtremeDB server) or remote (executing on different systems that are connected to the eXtremeDB server system via TCP/IP).  Systems can be heterogenous, e.g. a server on PowerPC and remote client on x86.

Scalable data management

An eXtremeDB database can be horizontally partitioned (“sharded”) for scalability. This feature leverages the processing power of multiple servers and/or multiple CPUs and/or multiple cores by distributing the data and the data processing to multiple instances of the eXtremeDB server. eXtremeDB’s distributed query engine isolates client applications from the details of the topology. When this technique is applied to multiple servers, the single-server limitation on the size of an in-memory database is overcome. In other words, if a single server has a maximum memory capacity of 768GB, then five servers have an aggregate capacity of 3.84TB.

eXtremeDB offers different distributed database options to address different objectives. This table summarizes the primary purpose and characteristics of each distributed database option (some of which may be combined, e.g. Sharding, Cluster and High Availability).  Learn more about distributed data management with eXtremeDB.