Attending the Teradata Partners conference and getting a briefing on the Teradata UDA – Unified Data Architecture. The UDA brings together Teradata’s traditional Enterprise Data Warehouse (EDW) products, its Teradata Aster discovery platform and a variety of partner technologies. UDA was announced in 2012 and they have been adding integrations and partnerships since then. Specifically they have:
- Improved connection to various Hadoop distributions
- Released a Fabric connector based on Infiniband
- Released SQL-H to give SQL access to Hadoop data
- Hadoop appliance and a strong partnership with Hortonworks to provide support on commodity hardware
- A new Extreme Data Appliance announced here at Partners 2013 designed to store data in the Teradata schema cost effectively
The whole UDA is heavily dependent on and integrated with partner products, reflecting the changing environment where multiple technologies and vendors must be applied to solve business problems.
Teradata report that more and more of their customers have integrated multiple elements (EDW, Hadoop, Aster) under the umbrella of the UDA. At Teradata Partners this year there are a bunch of customer presentations with several UDA adopters presenting their results in the last 12 months. In particular several customers integrating web clickstream data with their EDW data for customer analysis. Some are combining the EDW with Aster, some EDW and Hadoop, some all three. Clearly Teradata sees this as a direction for its largest clients, bringing Hadoop, traditional data warehouse and data discovery into a coherent architecture.
Aster is a key component in the UDA. Aster Discovery Platform 6 was recently announced (I blogged about the 5.x update of Aster recently) and brings three key elements:
- A new graph analytic engine using a SQL interface
Supporting complex graph-based analysis without requiring any expertise in graph technique mechanics
- A new data store, the Aster File Store
For storing the kind of data currently put into Hadoop in Aster
- SNAP Framework for discovery
Seamless Network Analytic Processing to bring the various engines and stores together so that there is a single, integrated SQL interface.
The graph engine provides an analytical and visual approach to analyzing data stored in standard forms – as distinct from requiring data to be stored in a graph format. The engine is designed to support SQL-based queries at scale, allowing graph analysis to execute scalably on existing data taking advantage of recent work by Google scalable graph analysis approaches. The use cases for this are well established – analyzing purchase patterns to find cross-over products, to find social networks for fraud detection or to see who might follow a customer or employee who quits.
The Aster file store is Hadoop-compatible and allows any Hadoop-like data to be stored in the Aster environment and accessed directly. SNAP meanwhile means that query optimization, workload management and more is integrated across the various data stores and query engines so that users can a common, single environment regardless. Users write their query and the SNAP engine makes sure it is executed and distributed appropriately given the data involved, how that data is stored and what kind of query is involved.
If you are interested in the whole issue of integrating various data types I have a blog post series on the impact of Big Data on Decision Management.