Objectivity ThingSpan

ThingSpan is a graph analytics platform that is targeted at data scientists, though it is worth commenting that Objectivity/DB is ACID compliant and it would therefore be reasonable to consider implementing the environment to support hybrid transactional/analytic processing (HTAP) as well as pure analytics. As can be seen from Figure 1 it leverages multiple open source technologies, with workflow support via DataFrames, and integration with both Kafka and Spark Streaming. Not shown on this diagram is the REST API.

Fig 01 – Thingspan leverages multiple open source technologies

As far as the database itself is concerned it is best to think of it is essentially a property graph with objects being equated with nodes and connections with relationships/edges. There are APIs for C++, Java (both with dynamic binding), Python and C# as well as a JSON importer capability. DO, which stands for “declarative objectivity”. Note that because Objectivity/DB is fully object-oriented there is no object relational mapping required.

ThingSpan is currently available on-premises, on AWS, and Microsoft Azure.

Customer Quotes

“Utilizing Objectivity’s data management technology, enabled MSC to implement a new level of productivity and user experience for CAE applications… We continue to work closely with the Objectivity product teams and the collaboration has been outstanding.”
MSC Software

“Shearwater holds a portfolio of proprietary technologies and in-house processing software enabling effective execution of geophysical surveys and delivery of high-quality data, and Objectivity’s technology is an important component of this geophysical platform, enabling the TRINAV Positioning System to handle the large volumes and complex positioning data that a vessel collects while at sea.”
Shearwater GeoServices

From a generic perspective ThingSpan is targeted at data scientists, particularly focused on Industrial (and defence and military) Internet of Things environments where there is a requirement for sensor fusion: where there are large numbers of sensors readings to be combined, in real-time, from multiple sources. Hence the support for Spark Streaming and Kafka to support real-time data ingestion. Support for machine learning is provide through support for Spark MLlib.

Fig 02 – ThingSpan Studio

Apart from its scalability and performance – no small matters in themselves – perhaps the two most important features of ThingSpan are the DO language and ThingSpan Studio. The latter, illustrated in Figure 2, provides browser-based visualisation capabilities both for exploration (notebook style), monitoring and “managed placement”, which allows you to cluster objects together for performance reasons, thus removing the need to do this within your applications.

DO, as its name implies, is a declarative language and the product has an appropriate database optimiser to support DO. More interesting is the fact that DO is SQL-like in its support for value-based queries (for example, SELECT name, location FROM School WHERE score > 90), Cypher-like for graph and navigational queries, and X-Path like for gathering reachable data. This illustrates the fact that the object-oriented underpinnings of ThingSpan effectively provide what might otherwise be called a multi-model approach. More generally, DO provides full CRUD capabilities for both the schema and the data (which can also be managed via ThingSpan Studio). It also supports the ability to query a graph for instances of a sub-graph (such as patterns of transactions that might represent fraudulent behaviour) and to compare such sub-graphs.

The key things about ThingSpan are scalability, performance and real-time capabilities, with connections being created as the data is ingested. It is this combination that makes ThingSpan stand out. There are other products in the market that focus on similar use cases to Objectivity but either they do not have the same degree of scalability and performance or they do not have the real-time explorative capabilities, at scale, that ThingSpan can offer. Similarly, while there are other multi-model database products these often require that you use different APIs to address the different storage models that are supported. For example, one such product requires four different languages to fully exploit it, whereas DO enables this through a single language. Moreover, we like the fact that this is declarative rather than procedural.

The Bottom Line

Objectivity/DB has proven scalability and performance credentials in highly demanding environments over a long period of time. ThingSpan inherits those characteristics while most other graph database products are parvenus by comparison.