EXtremeDB

History

McObject introduced eXtremeDB in 2001, targeting embedded systems running in resource-constrained environments (i.e. with limited random-access memory and relatively low-powered central processing units). eXtremeDB characteristics appealing to this market include a small code size (approximately 150 KB), native C language application programming interface, available source code, and a high degree of portability (to support the varied processors and operating systems used in embedded systems). Early deployments by customers included integration in digital TV set-top boxes, manufacturing and industrial control systems, and telecom/networking devices. eXtremeDB emerged to manage what industry analysts, and McObject, portray as significant growth in the amount of data managed on such devices; the company claims eXtremeDB was the first embedded database management system (DBMS) developed specifically for such environments

Later editions targeted the high performance non-embedded software market, including capital markets applications (algorithmic trading, order matching engines) and real-time caching for Web-based applications, including social networks and e-commerce. Features added to support this focus include a SQL ODBC and JDBC interfaces, 64-bit support, and multiversion concurrency control (MVCC) transaction management.

The increasingly popular term “Internet of Things”, especially since 2009, encompasses real-time embedded systems.

Core eXtremeDB engine

eXtremeDB supports the following features across its product family.

Application programming interfaces

Database indexes

Concurrency mechanisms

eXtremeDB supports multiple concurrent users, offering ACID-compliant transactions (as defined by Jim Gray) using either of two transaction managers: a multiple-reader, single writer (MURSIW) locking mechanism, or multiversion concurrency control (MVCC) transaction manager (optimistic non-locking model).

Supported data types

eXtremeDB can work with virtually all C language data types including complex types including structures, arrays, vectors and BLOBs. Unicode is supported.

Security

Distributed database management abilities

The eXtremeDB high availability edition supports both synchronous (2-safe) and asynchronous (1-safe) database replication, with automatic failover. eXtremeDB Cluster edition provides for shared-nothing database clustering. eXtremeDB also supports distributed query processing, in which the database is partitioned horizontally and the DBMS distributes query processing across multiple servers, CPUs and/or CPU cores. eXtremeDB supports heterogeneous client platforms (e.g. a mix of Windows, Linux and RTOSs) with its clustering and high availability features. A single partitioned database can include shards running on a mix of hardware and OS platforms

Hybrid storage

eXtremeDB Fusion edition provides the option of persistent storage (disk or flash) for specific tables, via a database schema notation.

Transaction logging

eXtremeDB Transaction Logging edition keeps a record of changes made to the database and uses this log to provide recovery in the event of device or system failure. This edition includes eXtremeDB Data Relay technology that replicates selected changes to external systems such as enterprise applications and database systems.

SQL ODBC/JDBC

The eXtremeSQL edition provides SQL ODBC support in eXtremeDB and a version 4, level 4 JDBC driver.

Kernel mode deployment

The eXtremeDB Kernel Mode edition deploys the database system within an operating system kernel, to provide database functions to kernel-based applications logic.

Features for managing market data

eXtremeDB Financial Edition provides features for managing market data (tick data) in applications such as algorithmic trading and order matching. A “sequences” data type supports columnar data layout and enables eXtremeDB to offer the benefits of a column-oriented database in handling time series data. The Financial Edition also provides a library of vector-based statistical functions to analyze data in sequences, and a performance monitor.

Benchmarks

McObject has published several reports on benchmark tests employing eXtremeDB to examine in-memory database system abilities. Main-Memory vs. RAM-Disk Databases: a Linux-Based Benchmark examined IMDS performance versus that of a traditional on-disk DBMS deployed on a RAM disk, on identical application tasks. The benchmark’s stated goal was to test the thesis that an IMDS streamlined architecture delivers a performance benefit beyond that provided by memory-based storage. Another benchmark, the Terabyte-Plus In-Memory Database System (IMDS) Benchmark, documented IMDS scalability and performance in the size range of large enterprise application (versus embedded systems) databases. For the test, engineers created a 1.17 terabyte, 15.54 billion row database with eXtremeDB on a 160-core SGI Altix 4700 system running SUSE Linux Enterprise Server 9.

In November, 2012 the Securities Technology Analysis Center (STAC®) released an audited STAC Report covering baseline STAC-M3 Benchmarks of eXtremeDB Financial Edition running with Kove XPD L2 storage on Dell servers with Mellanox InfiniBand. This configuration set records for lowest mean latency (response time) in 15 of the 17 benchmarks in the STAC-M3 suite of tests. STAC-M3 is designed for testing time-series management solutions (aka tick databases), which take historical and real-time streaming data as input, and perform user-defined operations on the resulting time series.

In late 2014, two additional audited STAC-M3 Benchmark reports focused on eXtremeDB Financial Edition. An October 29 report assessing McObject’s DBMS on IBM POWER8 hardware included new records in nine of the 17 benchmark tests. A November 18 report documented the first cloud-based STAC-M3, with eXtremeDB Financial Edition deployed on Lucera Financial Infrastructure’s Lucera Compute™ collocated cloud computing infrastructure for financial institutions and setting records in five of the 17 tests. The 2014 benchmarks with eXtremeDB were the first published STAC-M3 implementations developed in Python with SQL for database access

In 2016 the eXtremeDB Financial Edition was put through the STAC-M3™ Kanaga test suite, with results audited and published by the independent Securities Technology Analysis Center (STAC®). The STAC-M3 benchmarks consist of trading queries that were designed by firms on the STAC Benchmark Council to reflect the computing demands of real-world capital markets. Within STAC-M3, the Kanaga test suite’s much larger data set simulates actual data volumes encountered by capital markets systems. Each eXtremeDB mean response time was 5.5x to 212x faster than the previous best result.

Applications and customers