Teradata Vantage

Just recently I was at yet another vendor conference. This time Teradata in a not very sunny – cold and raining – Madrid. The main focus was on Teradata Vantage, which I had been briefed on previously. But you never really get all the nuances of a new release from just a briefing: the chance to talk in more detail to subject matter experts is always more revealing, and there’s a lot to like in Vantage.

The headline with Vantage is that the company has separated storage from compute. Teradata is hardly the first company to do this, but it has ramifications that I hadn’t previously thought about, which I’ll come back to in a moment. Before that, a couple of noteworthy points are that it will run on either your own hardware as well as on dedicated Teradata and, of course, in various options for cloud deployment. And, secondly, the product supports what the company calls 4D analytics and I would call dimensional analytics: meaning temporal, time series and geo-spatial capability. I’ll be producing some detailed research into this area in the coming months, so I’ll leave any discussion of this area for the time being, except to say that this is enabled by what was presented to us as “NewSQL”, being ANSI standard SQL with appropriate extensions. We (I wasn’t the only Bloor analyst at the conference) objected to this terminology on the basis that NewSQL already means something else. One of the executives we spoke to, who also didn’t like the use of this term, suggested that SQL+ might be better, but Couchbase is using SQL++ so that doesn’t work either. We recommended NextSQL. It remains to be seen what will actually be adopted.

Returning to the separation of storage from compute, the main reason usually touted for this is that you can scale one without the other (even, in the case of Vantage, in on-premises implementations), with all the cost advantages that that implies. However, what hadn’t really occurred to me was that also means that you can have different storage engines running against the same compute engines, and different compute engines running against the same storage engines and, of course, you can have many to many matches. Thus, Teradata has announced that it will be supporting Amazon S3 and Azure Blob storage engines and that it will be open to incorporating other storage engines in the future, as requirements dictate. On the compute side the company has already introduced both graph and machine learning engines in addition to its SQL engine. In the case of machine learning it is worth commenting that Python (for example) is converted into Teradata SQL, which means that you can parallelise Python-based query processes, something you cannot do if you run Python natively.

In effect, although Teradata is not positioning the product in this way, Vantage is a multi-model database. Along with the separation of storage and compute, the introduction of multi-model capabilities seems to be the major trend within the database world at present. And I am pleased to say that Teradata has implemented this properly. By this I mean that you don’t need to have multiple APIs (as you do with CosmosDB) to make it work.

My only caveat is that relational storage is not ideal for all sorts of analytics, even with S3 and Azure Blob, but the company’s willingness to consider linking to third-party storage engines could overcome any issues in this respect. Even bearing this caveat in mind, I am impressed. I feel like I have been waiting for this announcement for years, a bit like the anticipation of waiting for the last season of Game of Thrones or Avengers: Endgame. I may yet be disappointed by either of these, but I do not feel disappointed by Teradata Vantage.