The Role of Operational Analytics and Interoperability in the Era of IoT
The diversity and scale of the Internet of Things (IoT) will require many organizations to re-architect their data estates to enable support for multiple devices, data formats and new approaches to analytics.
There are distinct challenges to accomplishing this – in most instances this requires data to be ingested, curated and analyzed from millions of endpoints, with multiple data formats and multiple data models, at a high rate of speed. Meanwhile, there has been a strong drive to handle both operational and analytical needs in the same platform to reduce complexity – something we call ‘combined operational and analytic processing’ – in order to support operational analytics workloads driven by ‘multi-modal’ databases.
There are also new types of data platforms in the modern enterprise – SQL, NoSQL, Hadoop and so on – and a growing trend toward storing and processing data in multiple data formats in a single data platform, which we call ‘multi-model’.
These multi-model databases enable organizations to store and process data in multiple formats, enabling interoperability and providing the flexibility to evolve to keep up with the rapid pace of innovation involved in IoT, in terms of both changing data formats and the ability to apply the result of analytics processing to operational applications.
Current relational database technologies are challenged with delivering these benefits due to the inflexibility of fixed schema. However, SQL continues to be the default approach to analyzing data, so organizations deploying IoT need to rethink how they can optimize analytics for these deployments and consider multi-model, extremely scalable data platforms that combine support for SQL-based analysis with the flexibility to process multiple data formats and the agility to evolve to meet changing requirements.
Additionally, the scale of IoT projects will push computational and analytical capabilities closer to the network edge, where IoT devices live, in order to reduce the potential for the huge amount of data streaming from ‘things’ to saturate datacenter networks, storage and processing capabilities. Edge processing does not exist in isolation, and needs to be constantly updated based on analysis of the combination of real-time and historical data.