Teradata

As can be seen in Figure 1, Teradata supports a wide variety of languages for analytic purposes, most notably Python and R, though not Scala yet. Jupyter Notebooks are also supported. The company also offers an AnalyticOps Accelerator, a software and services offering comprised of best practices, proven design patterns, and tried-and-tested code, which jointly enable the client to implement an AnalyticOps Framework to support robust and continuous operationalisation of analytics. These assets have been taken from successful real-world projects and Teradata services, and genericised so that they may be reused and customised to other client settings. The services and technical IP of the AnalyticOps Accelerator are designed to help organisations get up and running with AnalyticOps and model management very quickly, accelerating deployment and increasing return on investment. While the company has told us that it intends to productise some or all of its AnalyticOps capabilities, it has not done so yet.

Figure 2 - Example of Teradata Vantage's 4D Analytics

As can be seen in Figure 2 Teradata also offers both graph capability and Advanced SQL, as well as support for machine learning.

As far as “Advanced SQL” is concerned, this represents Teradata’s extension to SQL that supports advanced analytic functions and machine learning at scale.  Time-series and temporal functions along with the company’s geo-spatial support are combined in what the company calls “4D Analytics”, as illustrated in Figure 2.

With respect to time-series, Teradata provides a Primary Time Index with data stored either by time, by column, or both. The data is organised into time buckets, or you can store it by, say, sensor ID. The former means that you don’t need to perform whole table scans. Round robin parallelism is provided, both to improve performance and prevent skew. Aggregation functions supported by Advanced SQL include statistical capabilities (mean, model standard deviation and so forth) as well as functions such as first, last, top, bottom and so on.

Alongside time-series and temporal capability, Teradata supports all the geo-spatial capabilities you would expect. Supported capabilities include points, curves and polygons, as well as functions such as measurements (distance, surface, perimeter), relationships (intersects, contains and so on), transformation and operation functions, and attribute abstraction such as the number of points within an area.

Figure 3 - Industry and analytic models offered by Teradata Vantage

Going beyond 4D Analytics, Teradata also offers a wide range of industry and analytic models. Some examples of the latter are shown in Figure 3.

As can be seen in Figure 1, Teradata supports a wide variety of languages for analytic purposes, most notably Python and R, though not Scala yet. Jupyter Notebooks are also supported. The company also offers an AnalyticOps Accelerator, a software and services offering comprised of best practices, proven design patterns, and tried-and-tested code, which jointly enable the client to implement an AnalyticOps Framework to support robust and continuous operationalisation of analytics. These assets have been taken from successful real-world projects and Teradata services, and genericised so that they may be reused and customised to other client settings. The services and technical IP of the AnalyticOps Accelerator are designed to help organisations get up and running with AnalyticOps and model management very quickly, accelerating deployment and increasing return on investment. While the company has told us that it intends to productise some or all of its AnalyticOps capabilities, it has not done so yet.

As can be seen in Figure 2 Teradata also offers both graph capability and Advanced SQL, as well as support for machine learning.

As far as “Advanced SQL” is concerned, this represents Teradata’s extension to SQL that supports advanced analytic functions and machine learning at scale.  Time-series and temporal functions along with the company’s geo-spatial support are combined in what the company calls “4D Analytics”, as illustrated in Figure 2.

With respect to time-series, Teradata provides a Primary Time Index with data stored either by time, by column, or both. The data is organised into time buckets, or you can store it by, say, sensor ID. The former means that you don’t need to perform whole table scans. Round robin parallelism is provided, both to improve performance and prevent skew. Aggregation functions supported by Advanced SQL include statistical capabilities (mean, model standard deviation and so forth) as well as functions such as first, last, top, bottom and so on.

Alongside time-series and temporal capability, Teradata supports all the geo-spatial capabilities you would expect. Supported capabilities include points, curves and polygons, as well as functions such as measurements (distance, surface, perimeter), relationships (intersects, contains and so on), transformation and operation functions, and attribute abstraction such as the number of points within an area.

Going beyond 4D Analytics, Teradata also offers a wide range of industry and analytic models. Some examples of the latter are shown in Figure 3.

Fig 02 Teradata Advanced SQL support

By 4D Analytics Teradata means geo-spatial capabilities, which are three-dimensional, plus time, which is the fourth dimension, with time-based support including both time-series and temporal capabilities. Advanced SQL represents Teradata’s extension to SQL (for example, GROUPBY TIME) to support the 4D functions shown in Figure 2.

Starting with time-series, Teradata provides a Primary Time Index with data stored either by time, by column or both. The data is organised into time buckets, or you can store it by, say, sensor ID. The former means that you don’t need to perform whole table scans. Round robin parallelism is provided, both to improve performance and prevent skew. Aggregation functions supported by Advanced SQL include statistical capabilities (mean, mode, standard deviation and so forth) as well as functions such as first, last, top, bottom and so on.

Temporal data differs from time-series data to the extent that temporal data is fixed for a particular period of time. Temporal data has a start and end date, even if the latter may be putative rather than actual. The most obvious use cases include demographics and many things associated with human activity, for example, tracking and analysing emergency vehicle response times to do trend analysis, or in logistics. Auditing (and not just financial auditing) and compliance are other environments that lend themselves to temporal analytics, especially where you need to look backwards in time. For example, when you need to distinguish between transaction time and valid time. In analytic environments a further example would be when a product, say, is moved from one product group to another. Teradata supports all of these use cases. Competitive offerings may be able to support the product change scenario via slowly changing dimensions (which you can do with Teradata also) but cannot support the “point in time” queries that 4D Analytics supports.

Finally, Teradata supports all the geo-spatial capabilities you would expect. Supported capabilities include points, curves and polygons, as well as functions such as measurements (distance, surface, perimeter), relationships (intersects, contains and so on), transformation and operation functions, and attribute abstraction such as the number of points within an area.

As one might expect you can use any of the leading business intelligence, visualisation, and analytics tools, as well as notebooks and IDEs along with 4D Analytics.

Fig 03 Example of bi-temporality

Time-series analytics are essential for many predictive and prescriptive analyses as well as such things as fraud detection. Many of these use cases – especially in IoT and the industrial sector – also require geo-spatial capability to be combined with the time dimension. Temporal processing allows bi-temporal queries – for example, see Figure 3 – in a user-friendly fashion that doesn’t require knowledge of effective data fields and/or complex condition clauses.

There are plenty of vendors that have geo-spatial capabilities. There are some that have support for time-series, though only a minority of these offer sophisticated analytic functionality. There are a few that have temporal capabilities though all of these except for Teradata are primarily operational/transactional databases rather than analytic. What that adds up to is that Teradata Vantage is the only product that we are aware of that has all the 4D capabilities that you might need. Perhaps the one caveat that we would make is that if you only want the sort of DevOps metric and monitoring solution that some time-series vendors target, then Teradata Vantage would probably be overkill. But if you want to combine that sort of solution with anything more sophisticated then Vantage would certainly be appropriate.

The Bottom Line

Of course, Teradata Vantage covers much more than the 4D Analytics we have discussed. While it has significant advantages when it comes to time-series and/or temporal processing the product has lots of other good things going for it. It should pretty much be on everyone’s shortlist when time-series analytics is at issue.