We’ve been working on the next major release for Ajilius for several months now, and it is time to let you know what we’ll deliver starting next month.
Q: “What is the fastest algorithm to load a dimension?”
A: “It depends.”
That’s right, the best algorithm for a load process depends on your platform and your data. MPP platforms prefer CTAS algorithms, SMP may do better with UPDATE algorithms for low volatility tables.
What we often see, though, is that design decisions made during development are no longer valid when the DW has been in production for several months. Assuming good design choices in the first place, the most common reason is that the volume and patterns of data in production are not matching those that drove the initial design decision.
Ajilius already supports a choice of CTAS or UPDATE algorithms depending on the platform, but in Ajilius 3 we’re taking an industry-leading leap forward and introducing a dynamic choice of algorithm at run time. We’re calling it Adaptive ELT.
Based on the combination of platform and data, we’ll make an intelligent choice about the best algorithm to use for each and every batch run. Does the data suggest you’ll be updating the majority of rows in a billion row table? We’ll generate CTAS for that batch. Are you only updating a few rows on the next run? We’ll use an UPDATE statement.
No other ETL or ELT platform gives you this flexibility. It is like having your developers rewrite and tune their code, each night, with the data for that night’s batch driving their decisions. You’ll get an expert ELT system that dynamically monitors and adjusts your processing for maximum performance, on every single run.
Until now, Ajilius has generated scripts for ELT processing. Ajilius 1 generated Python, Ajilius 2 generates Groovy, Ajilius 3 is throwing away the single script constraints.
In Ajilius 3 we’ve separated generation of SQL from generation of the wrapping script. We’re calling this SQL+Script, because we’ll be generating SQL for every transform, supported by a wrapping Script of a type requested by the user.
A new template option will enable the generated SQL code that effects transformations to be wrapped in multiple scripting languages. We’ll continue to support Groovy as our primary scripting option, but here are just some of the options that will be available in Ajilius 3:
- Plain SQL. Every script will be able to be generated as just SQL statements in the target platform dialect. This will delight customers who want to use Ajilius in a hybrid environment with existing data warehouses.
- Stored Procedures. SQL Scripts will be wrapped in stored procedure creation scripts for customers who prefer to use tools like SQL Agent and SSIS for job execution.
- PowerShell. The rise of PowerShell as an automation language, and its growing popularity in CI/CD environments, makes it a great option for an Ajilius target. We’ll generate wrappers that emulate our current Groovy scripts, but using PowerShell as the scripting language.
Being great fans of Lawrence Corr’s Agile Data Warehouse Design methodology, we’re adding BEAM support to Ajilius.
Our new “Model” module will implement the BEAM methodology. Here’s what you’ll see in the first release:
- Capture and presentation of 7-Ws
- Generation of template fact/s and dimensions from Ws
- Mapping of staging tables to fact/s and dimensions
- BEAM-compatible short codes for all object types.
- Incorporation of BEAM artefacts into Ajilius-generated documentation
Ajilius currently works “left-to-right”, in that we start from data sources and eventually transform to facts and dimension. We’ll keep that capability in Ajilius 3, but we’ll also provide “right-to-left” model driven implementations, where you’ll have the option to create facts and dimensions first, then link them back to transformations. In other words, we’ll support model-driven, data-driven and hybrid approaches to building a dimensional data warehouse.
All of this is an extension to the current Ajilius. You’ll be able to retrofit BEAM to an existing Ajilius DW, as well as integrate new BEAM-designed stars in to a conformed set of existing dimensions.
Once again, Ajilius is demonstrating innovation and leadership in data warehouse automation, being the first to directly integrate BEAM into the automation process.
The Fine Print
We’ll be rolling out the first release of Ajilius 3 some time before June 30.
The Ajilius 3 upgrade will preserve all your existing metadata, but we’ll be making some adjustments to the metadata structure that will break the incremental update flow. In other words, we’ll be expecting all users to upgrade to 3, then restarting incremental upgrades from that point.
We will continue to support Ajilius 2 for six months from the date of release of Ajilius 3, in order to give users time to evaluate and test the migration process.
There is, of course, no additional charge to users for Ajilius 3, all upgrades are covered for the life of your license agreement.
We hope you’re as excited as we are by these changes!
Ajilius. Committed to innovation in data warehouse automation.