Teradata is on a Tear

developmentFor a company that many competing vendors want to write off as “last gen” analytics, Teradata has two things going for it. First, Teradata customers have little interest in throwing out the enterprise-grade data warehouse baby with the Hadoop data lake bathwater. Second, the company is incredibly active in making the transition to the modern big data world.

Here’s my take on the many Teradata announcements in the last couple of months:
  1. Development of Presto. If you haven’t been following, Teradata decided to be the corporate sponsor for Presto in an effort to provide a robust and mature SQL-on-Hadoop option. Well, not only is Teradata executing on that promise, but it is also listening to customers’ feedback and has accelerated ODBC/JDBC drivers to be delivered this year. This type of enhancement may not be strictly open source, but it is free for use.
  1. Teradata for DevOps. The company has created a Python module and the ability to wrap scripts around common data warehouse activities. This may make it easier and more familiar for “DevOps-style” application building and connections, and has already proven popular as open source on DevX and Pypy with 1,000 downloads in the first week. I say “DevOps-style” because I’m not convinced the users quite fit the label, but it’s a great start anyway. 
  1. Analytics in the cloud. Teradata is moving with the times around consumption models and preferences, going from software to appliances to its own cloud offering—and the coming availability of the Teradata database in AWS for production environments (not just testing and trial as currently) in Q1 will be appreciated. If a Microsoft Azure offering follows later next year, well, this all makes Teradata technology more consumable and easier to access. I’d like to see migrations and hybrid environments to/from/between clouds, but the official answer is “not yet.” 
  1. Internet of Things/Analytics of Things. New Listener software is beta now with an expected GA in Q4. Teradata is taking a black box approach, not DIY assembly for moving data, which uses Kafka, Cassandra, Mesos, ElasticSearch, Openstack, Docker, and APIs, but all pre-integrated. Supporting real-time data ingestions from multiple sources with intelligence in pipeline and the ability to write streams to multiple targets is goodness, even as it competes with Hortonworks Dataflow, AWS Kinesis, etc.   
  1. Aster Analytics on Hadoop. Teradata wants to bring all their powerful Aster analytics functions to sensor data with a YARN enabled product for Hortonworks and Cloudera that should be GA on their appliance in Q1 and GA for commodity Hadoop clusters in Q2. An important caution is that Aster on Hadoop will likely be much slower, but the economics mean you can add more nodes. Over time, the Aster appliance may then fade out of the picture. 
  1. Unified Data Architecture (UDA). Teradata is making the UDA simpler, faster, and more versatile with integrated QueryGrid and Presto, and SQL as common protocol or interconnect language, all as an alternative to Hive. This allows you to have access from Hadoop to Teradata and vice versa. It also enables Teradata to act as a connector to Presto (and Presto then to Hadoop, Cassandra, MySQL, and Postgres). They are also offering a new UDA appliance with Hadoop, Aster, and Teradata in a single rack. If Oracle’s success with their Big Data Appliance (BDA) can be replicated, this plays well to the customer who is comfortable with hardware-based data warehouses and just wants to extend and optimize in the same fashion.

Beyond all this, there was more news around hardware platform updates, including the 6800 for EDW (shipping now) and the 1800 for big data, which can be built to suit the workload profile desired. Not least, Teradata’s Think Big managed services for Hadoop is playing as “unicorns for hire” for outsourced data scientists, platform and application support, and general staff augmentation.

A whole lot going on for a company others would like to disrupt, eh?


big data analysis