I saw a presentation on big analytics appliance which gives Hadoop, aster and teradata all in one appliance. and then a language called sql-h which allows query on all 3.
A question I had was that the load on the hadoop grows very fast because people pour in a lot of data which is not used very frequently. (at least in my case).
Is it possible that if I have to just grown the hadoop nodes then I install the teradata hadoop distrubtion on a commodity server (cheaper normal server from HP/Dell) and then add it to the cluster?
or should all nodes be the teradata hardware?