Teradata Aster

Teradata Aster

Yours truly spent a couple of hours recently at the Teradata Aster Manchester office with Vic Winch, Director of the Teradata Aster COE, EMEA. Vic kindly offered to bring VLDB up to speed on all things Teradata Aster. We love a good geek-off, so how could we possibly refuse?

Vic an I have been bumping into each other since the late 1980's. Vic worked for Teradata way back then, and I was a graduate trainee at Royal Insurance (now RSA Group) in Liverpool. Just like Littlewoods (now ShopDirect) over the road in Old Hall St, Royal Insurance were very early Teradata adopters. Two Teradata systems in the same street in central Liverpool in the 1980's - imagine that!

Aster has been doing POCs in EMEA during 2012 and 2013. This has covered various industry sectors, and not just the usual suspects in finance, retail and telecoms. Some of these POCs have been successfully converted into customers.

The current Aster go-to-market strategy, for EMEA at least, is appliance only. The cloud and software only models bring too big a support headache. For folks that want to kick the Aster tyres before moving to a POC, the Aster virtual machine (VM) can be downloaded here: http://www.asterdata.com/downloads/

A typcial entry level Aster appliance consists of 4 Aster nodes (2 queen nodes and 2 worker nodes) giving 10TB of space, and 5 Hadoop nodes (2 name nodes and 3 data nodes) giving another 30TB of space. All of this ships as a Teradata Aster appliance with all the servers, storage, networking, monitoring services and support that Teradata customers currently enjoy. Pricing seems to be similar to the Teradata data warehouse appliance platform range.

So far we have a Teradata Aster appliance with the PostgreSQL derived parallel Aster DBMS on one set of nodes, and the Hortonworks version of Hadoop running on a separate set of nodes. This brings the usual pros and cons of each: SQL ease-of-use but lack of analytic features for the DBMS; Java + MapReduce complexity and feature richness for Hadoop. Ho-hum.

Fear ye not. This is where the Aster 'secret sauce' really delivers. The Aster SQL-H interface gives the best of both worlds: SQL ease-of-use and Hadoop MapReduce functionality. Not only that, the nice folks at Aster provide a comprehensive library of ~70 functions so that mere mortals don't have to write MapReduce programs. Back of the net!!!

For those that don't understand the value of this approach, trying writing some set based SQL to analyse large volumes of transaction data in date/time order. Then try it on an MPP system where the data is physically distributed across an arbitrary number of nodes.  Not pretty or efficient. A much simpler way would be to invoke the request from SQL, which we all know and some even love, and for a pre-written MapReduce procedure to do all the hard work for us on the Hadoop stack.And that's not all folks. Once we have the output from our analysis, Aster also provides a data visualisation capability as part of the offering. This is another component in the Teradata Aster Discovery Portfolio.We'll report back on Aster once we've had a play with the VM. It certainly looks *very* interesting.

Once again, many thanks to Vic Winch and Teradata UK for their hospitality :-)