Teradata Climbs Up the Stack with ‘Analytics Platform’ Strategy
Teradata is extending the types of data, algorithms, and frameworks it can store and run in its flagship MPP database as part of the new Teradata Analytics Platform unveiled at its annual user conference today. Shipping early next year, the new platform is being billed as the next iteration of the enterprise data warehouse, and a one-stop shop for all of its customers’ analytic needs.
For most of its history, Teradata has focused its efforts on building core underlying data storage and management functions into its massively parallel processing (MPP), column-oriented relational database. With that foundation in place, customers could then write their own analytic routines using SQL, run pre-generated SQL from business intelligence tools, or run packaged analytic applications from the likes of SAS and other vendors.
The company revealed a desire to go “up the stack” in 2011, when it spent $230 million to acquire Aster Data Systems. Aster had developed SQL- and MapReduce-based software that let customers analyze large amounts of semi-structured and unstructured data, such as Internet clickstreams and social media activity, for the purpose of generating product recommendations, detecting fraud, and reducing churn.
While Aster helped Teradata break into the burgeoning market for big data analytics, the Aster technology hasn’t been fully integrated with the company’s traditional EDW business. There were integration points, such as the QueryGrid functionality Teradata unveiled in 2014, and its foray into pushdown processing. But the two products largely lived separate lives.
That separation will be erased when Aster technology is fully ported to run within Teradata’s EDW. That’s part and parcel of the Teradata Analytics Platform plan, says Imad Birouty, Teradata’s director of product marketing.
“We’re actually moving Aster and Teradata together. This is step number one in…bringing those analytics over,” Birouty tells Datanami. “We’re taking those advanced analytic functions that were in Aster, and we’re porting them over to the Teradata database. So those will be embedded directly in the database.”
The plan calls for having the first batch of Aster functions — including path analysis, attribution, text analysis, sessionization, and time series analytics –available in version 16.20 of the Teradata database in the first half of 2018, Birouty says. A second batch of Aster functions, including graph functions and other capabilities, are slated for an update that ships in the second half of the year, he adds.
“[Aster] was multi genre. It did graph functions. It did machine learning. It did text analytics. It did a lot of things, and so we’re bringing all those over, little by little,” Birouty says. Eventually, all of Aster’s functions will be ported to run under the Teradata database, and Aster will cease to exist. “It’s going to take a couple of years to do that.”
About a year from now, the Teradata EDW will be updated to natively support other analytic frameworks, including things like Apache Spark, Tensorflow, Theano, and even Gluon, the new machine learning library recently unveiled by Microsoft and AWS. “We’re looking at a lot of the popular tools out there and whatever is the hottest thing, we’ll plug them in,” Birouty says.
These frameworks will be brought into Teradata by way of a “pluggable architecture” that the company is building into its Teradata Analytics Platform, according to Birouty. When it’s ready, the architecture will be transparent to the user and work entirely “under the covers,” he says. However, that system isn’t quite cooked. “We’re not there yet,” he adds.
Teradata is seeking to be just as open when it comes to languages like R and Python and tools like SAS, KNIME, RStudio, Jupyter, and Dataiku, the company says. The database warehouse already had support for JSON data types, and support for Avro and other data formats will be added.
Centralizing the data and the analytics in the Teradata EDW will bring several benefits to customers. For starters, it will minimize the time and expense of moving data over the network. It will also eliminate the need for the IT department to maintain additional systems.
Co-mingling different data sets and data types will also allow customers to experiment with new analytic techniques, Birouty says. “You can imagine you can do some pretty amazing things if you have all these analytics together,” he says. “You can analyze the data that’s in your warehouse and IoT data together.”
For example, a trucking company could use sensor data to track the movement of its trucks while also monitoring the content of the cargo or special business relationships. If the IoT data indicates a delivery to a particular customer is going to be late, while the EDW data shows that the cargo is sensitive or the customer relationship has a special attribute, the trucking outfit could take pains to expedite the delivery or alert the customer to the late delivery.
“This is pretty much unprecedented in the capabilities that it’s bringing together, all at the fingertips of the user,” Birouty says. “The users don’t have to worry about separate pools. All the analytic functions, the tools they have, the data types, are all supported within this Teradata Analytics Platform. You can operatize the insights and deploy them to all the users. Pretty powerful stuff.”
Rather than a dramatic course change, however, the unveiling of the Teradata Analytics Platform represents an extension of the company’s previous strategy, Birouty says. “We refer to the future of the data warehouse as the analytics platform,” he says. “So the Teradata Analytics Platform is a superset [of capabilities] — the next generation with more capabilities than just the data warehouse today.”
Teradata also announced IntelliSphere, a prepackaged set of tools often used in data science and data analysis work. “We put together a software package … that are all the products you need to put together your analytic ecosystem,” Birouty says. “All the software you need to ingest data, to mange it, to deploy apps within the environment, to access the data.”
The company made the announcements today at Partners, its annual user conference, which is taking place this week in Anaheim, California.