I've got a slightly bigger problem. How to archive a Teradata appliance.

The solution I've gone with is Hadoop. It ain't cheap but it is cheaper than having to expand the Teradata box. It is actually buying us time to consider how to get the best out of Teradata.

Much of the time an archive is there "just in case". Quite often when you dig into the business requirements for an archive solution it is remarkably sparse. There are few, if any performance NFRs or RTO, RPO criteria. The stated requirements could be summed up as store data securely for 'n' years and be able to bring it back in the event of a compliance query.

With that level of requirements where is the onus on me to store the data in expensive high performance equipment?

Then throw into the mix that the few queries that are turn against archive data are the sort where a slow running batch process is perfectly acceptable. MapReduce (or more accurately HIVE SQL query) = slow running batch process.