Teradata applies time series analytics tools to IoT data
Data warehouse pioneer Teradata looks to ease IoT data analysis with capabilities that address skills gaps on time series analytics techniques, which are coming more to the fore.
The internet of things continually emits a torrent of data, which usually must be analyzed in terms of when it was created to be of full value. IoT data analysis often calls for specialized time series methods to be applied. But while they're well-known in areas such as signal processing, economics and weather forecasting, there is a gap in such expertise in many enterprises.
Data warehouse pioneer Teradata is looking to fill in that sensor analytics gap with new capabilities for its Teradata Analytics Platform. In a new update, existing geospatial data analytics tools are enhanced with time series analytics capabilities tuned to deal with the types of problems that IoT data creates. While time has always been a part of most business analysis, the Teradata platform seeks to deal with the newly arriving IoT data in a new way.
"IoT data has very different characteristics than traditional enterprise data and even other big data sources," said Barry Devlin, founder and principal of 9sight Consulting in Cape Town, South Africa.
Barry Devlin
Teradata has worked to bring time series structures into its tools in such a way as to make it easier for SQL users to work in the rarefied space of time series data, Devlin said.
He said IoT data can be very "dirty," in the sense that values may be missing or in error in various ways. "Data structures may change unexpectedly, for example, when a sensor is replaced in the field," Devlin said, adding that better skills, tools and techniques help in handling those issues.
IoT breadcrumb trail
The Teradata platform update helps users deal with missing data, incorrect values and other anomalies in several ways, according to Imad Birouty, director of technical product marketing at Teradata, based in San Diego.
The system can lay out the data for quick access and processing, handling anomalies and missing values automatically, Birouty said. For example, it can deal with the varying frequency rates at which sensors report data; that can be 10 times per second in some cases and 10 times per hour in others, he said.
To help make that possible, a new primary index provides a way to distribute time series data quickly into specified segments of time. Related time series analytics is further supported by query methods optimized for such applications, Birouty said.
Time series analytics provides, in Birouty's words, "a trail of breadcrumbs that let you re-create what happened" in areas of IoT data analysis.
Time is of the essence
Teradata's move is part of a wider trend that sees data platforms and analytics software adapting to the needs of IoT users.
For example, Siemens Healthineers recently said it was using SAS IoT analytics software to study data generated by its magnetic resonance imaging and computed tomography systems to predict system problems and downtime.
Meanwhile, startup Swim.AI this month exited stealth mode, releasing its Swim EDX system for applying analytics on remote -- or edge -- devices, with the goal of reducing the load on central data warehouses.
IoT brings all aspects of time to the fore.
Barry Devlinfounder and principal of 9sight Consulting
And machine data specialist Splunk said it would later this month roll out a limited release of a Splunk Industrial Asset Intelligence platform for IoT data analytics uses.
According to analyst Devlin, the need for reliability in IoT applications may bode well for software such as the Teradata Analytics Platform, which is largely centered on established relational database technology. That is despite the new time series analytics wrinkles built into the platform.
"The bottom line is that time has always been an important, but underappreciated, factor in data warehousing and analytics," Devlin said. "IoT brings all aspects of time to the fore."
Barry Devlin
