Ataccama

Fig 02 - The Ataccama ONE Knowledge Catalog

The other notable result of the deep integration provided by Ataccama ONE is that the various user interfaces that are provided, which are targeted at different personas, are consistent across all of the platform’s capabilities. As illustrated in Figure 2, this is visually appealing. More notably, this screenshot shows the forthcoming release (2.0) of Ataccama ONE, in which the company is introducing its Knowledge Catalog as a superset of both its Data Catalog and Business Glossary.

More generally, the company describes its platform as “self-driving”, by which it means that it has built significant levels of machine learning and AI into Ataccama ONE. As an example, the data catalogue, which includes full text search and supports natural language processing, will learn from past experience of relevant searches, and come up with notifications if it detects an anomaly or other alert-worthy information, such as “looks like a table with GDPR content”. While on this topic, it is worth noting the company’s support for data masking (both static and dynamic), which includes consistent masking and the ability to preserve referential integrity, amongst other algorithms. In addition, consent management is provided within the MDM module. Unfortunately, there are no features to support the automation of data subject access requests (DSARs). While the company is by no means unique in building machine learning into its platform we particularly like the fact that it has also built explanations of its AI into the offering. We have not seen this level of explainability from other vendors.

Other features of note include the fact that the IDE is Eclipse-based, there is support for REST APIs and, where relevant, jobs can run natively on Spark, Databricks and so on, so that transformations can either be supported in ETL or ELT mode. Beyond this, the company provides a Reference Data Management module, with approval-based workflow capabilities and governance workflows built in to manage and distribute reference data. Data quality rules can be exposed as a web service.

Ataccama ONE is a comprehensive platform for establishing and managing (cloud-based) analytic environments. It has all the fundamental requirements that one could ask for, though from a marketing perspective it might be advantageous to the company if it called out data governance as a module in its own right. That said, spreading governance capabilities across the environment makes a lot of sense so this is a marketing argument rather than a technical one.

Ataccama ONE has the distinct advantage of being a natively integrated, single engine solution, as opposed to a bunch of individual modules that have been stitched together. This is not only relevant when it comes to the performance and scalability of the platform but also to the end-user interface, which is consistent – by persona – across the environment. The product’s “self-driving” approach to automation is also significant and perhaps ahead of some of its competitors. That said, all major market players are investing into AI and automation capabilities, and one area where Ataccama stands out is in its explainability.

The Bottom Line

Ataccama, while by no means one of the biggest names in this space, is a significant contender with a proven track record. It is certainly worth serious consideration as a potential supplier.

Ataccama’s data fabric solution works by first ingesting the metadata within your system. This metadata can be gathered automatically via a discovery process, imported, or created manually. For the former in particular, this involves examining your data assets and inferring metadata from them within Ataccama’s data catalogue. Finding and making explicit the relationship between two assets is one example of this; detecting the business domains associated with each data asset would be another. In either case, data profiling and classification – provided by the platform – are employed as part of this process.

Moreover, Ataccama is built on top of a metadata knowledge graph. This means that when a user or process requests some data, the platform will extract metadata pertaining to that use case, as described above, then compare it against the metadata in said knowledge graph in order to decide which data to return and how to process it (via aggregation, transformation, filtering, and so forth). In fact, this graph comes with an AI-driven decision framework for just this purpose. Notably, the graph is easily human-readable, meaning that it can also be useful for viewing and understanding the connections within your system.

Following that, various checking and monitoring functionality is applied to assess metadata integrity, compute data quality scores (including domain-level aggregate scores), detect anomalies, and so on. Note that everything from metadata ingestion to this point is fully automated, but still transparent to the end user. The pipelines can now, optionally, be interacted with manually for the purposes of data cleansing, master data management and such, then finally they are used to deliver the now curated and remediated data to the consumer.

In addition, the data pipelines that Ataccama builds are composites: there can be comprised of several different technologies assembled together into a single pipeline. In other words, they have a composable architecture. This is important: data fabrics need to stretch across your entire environment and coordinate with everything therein. Since Ataccama, or any other data fabric provider for that matter, cannot possibly offer every piece of functionality that might need to be in a data pipeline – nor would you necessarily want it to – this sort of composability is essential for creating a robust data fabric.

Fig 2 - Example Ataccama data fabric architecture incorporating data quality

Data fabrics in general, and Ataccama in particular, effectively provide enhanced data integration through the means of massively more automated data flows. At least in Ataccama’s case, this increase in automation applies in terms of scope (far more of the pipelining process can be automated), speed (pipelines are created and executed more quickly) and accuracy (pipelines deliver more relevant data more often). In particular, Ataccama can automate the governance and management of your data (including, say, data quality – see Figure 2) by building those capabilities into the pipelines created by your data fabric.

It is also worth noting that Ataccama’s offerings in general have historically been metadata-driven, making the data fabric a rather natural evolution of the company’s existing strategy. This can be seen in, for instance, its solution for master data management, which does not form a data fabric per se but is certainly metadata-driven, and benefits substantially from it. In other words, Ataccama has been using the techniques core to its data fabric for most of its existence – perhaps even longer than the term itself has been around for – giving it substantial experience with, and time to develop, those techniques.

The Bottom Line

Ataccama provides a suite of data management products within the ONE platform that can quite readily come together to create a data fabric. If you’re looking for a company to help you create your data fabric, Ataccama’s experience with the techniques that can drive them make it a capable, and safe, pair of hands.