TransLattice: TED Redefines the “Distributed” in Distributed Database
With all the early stage and open source not-strictly-relational database vendors spreading over the IT landscape, many with an eye towards Web 2.0 OLTP and Big Data, it becomes difficult to differentiate. We find plenty of new NewSQL options, and an even longer list of NoSQL options in its many variations (Graph, Columnar, document-oriented, etc.). I refer you to the reference work over at nosql-data.org for about as an exhaustive list as you are likely to find.
TransLattice took a clear step in differentiating itself from the no/NewSQL database horde with its release of TED on July 24, 2012. TED, or TransLattice Elastic Database, a NewSQL variant, redefines the notion of "distributed database" from a network-distributed database historically measured in feet, meters, or over campus distances to a database with potentially a worldwide geographic distribution footprint. TED could be considered a lower-cost alternative to highly available, highly scalable RDBMS offerings like Oracle RAC and IBM DB2 pureScale.
TED, primarily designed to support highly available geographically dispersed OLTP applications, runs in VMware instances and appliances thereof, or as a Cloud service, or in hybrid mode. It lacks the notion of a name node to avoid single-point-of-failure syndrome and supports ACID requirements by making intelligent use of replication. TED leverages Postgres at its core though clearly TED is fully forked from the original open source object-relational database. With TED you fire up nodes all over the world, assuming the proper infrastructure is in place, inside of an hour, but the resulting logical database looks like a single, local database to developers and users.
Some of the primary design considerations include:
- Policy-based: Data is routed and replicated based on policies, with the goal of matching data location as close as possible to the usage network edge. The policy model is carried-out by multiple virtual administration engines working together through a "global consensus protocol."
- Machine-learning: TED is adding machine-learning so that it can enhance policy-based data distribution by analyzing network traffic.
- Local optimization and availability: Though not a strict requirement, TED implementations do best when two nodes are deployed in the same location to support monitoring, availability, scalability, and data protection requirements at the edge.
- Fault tolerant: Even in situations where a node or node combination goes offline, administration kicks in and routes data to the next most optimal node(s) - user visible latency may temporarily increase during the outage, but the database continues to function. When the outage has been remedied data traffic returns to the optimal pattern. TED works with logging servers like Splunk in order to fit into larger systems management solutions.
- Scalable: Nodes can be added or substracted without interruption.
TED represents three years of research and engineering, and is largely the brainchild of CTO and co-founder Mike Lyle, who was convinced distributed database design could stand an overhaul. The trick to making TED a truly geographically distributed database involved: Ripping and replacing the native (1) storage engine and (2) query processor of Postgres, and (3) adding the intelligent policy management. While TED certainly was designed with Hadoop MapReduce in mind, ESG sees deployments focusing on large OLTP opportunities, starting with the usual vertical suspects for groundbreaking technology, financial services and government.
TransLattice, like many early stage database companies, will lean TED go-to-market mainly on executive committee connections to establish an initial customer base. ESG suspects, however, that the unique aspects of TED will drive a healthy amount of voluntary, inbound leads.