June 29, 2020The 1990’s Called, They Want Their OLAP Back
In the modern day world, data is created with every step you take. Buying coffee with your credit card? Data. Navigating with Google Maps? Scrolling Instagram? Data! With the world becoming increasingly data-driven, the more you know, the better you can make business decisions. From sales analyses to marketing effectiveness; from inventory optimization to Customer360; every enterprise is trying to find ways to better understand all angles of their business.
Why is OLAP important
With oceans of data at our hands, how do we make sense of all that we’re collecting? How do we answer questions like:
- Sales growth between this month and the same month last year?
- What are the primary demographics for each of our product lines?
Data is typically collected in rows and columns, but in order to convert these numbers into useful insight, we rely on systems and processes that provide aggregations for analytics. This is where OLAP comes in; OLAP stands for Online Analytical Processing and is the backbone of business analytics. It is the process to organize the “Facts”, which are the raw numbers, and aggregate them across different angles, which are referred to as “Dimensions”. For example, your fact may be the sales amount of each transaction, and you may need to find the Total Sales number sliced by dimensions like Time or City. These dimensions can also have multiple levels, which allow users to perform drill downs like from Country to State to City to Zip Code, etc.
Different flavors of OLAP
While OLAP provides users with flexibility, there are several approaches to serve the information they need.
Multidimensional OLAP (MOLAP) is the traditional classic form. The users pre-select a subset of dimensions and facts, the system builds a data cube, then users can query the cube. While this approach can provide quick query responses once the cube is ready, it’s only sustainable when there are not too many dimensions in the cube because it requires a lot of upfront computation of every iteration and intersection even if many are never going to be queried by users. Every time there are changes made to the cube, it has to be completely re-computed, which can take days. Another issue with this approach is the “data explosion”. When every combination of elements and dimensions have to be computed against all measures, the resulting data can be magnitudes greater than even the raw data itself.
The second approach is Relational OLAP (ROLAP) which gives users OLAP-like capabilities in the data stored in the underlying database. This approach allows users to run sql queries with slicing and dicing. The advantage is that data movement is eliminated and architecture is simplified, but the performance of the queries are often slow.
A third approach is the new Cloud OLAP or COLAP which makes OLAP work as a service at Cloud scale to analyze large amounts of data without moving it out of customers’ cloud data warehouses or data lakes.
AtScale takes the better parts of the DNA from each approach and splice them with machine learning. Start off with a virtual ROLAP model, our approach allows users to run reports right away in an OLAP manner. Then the AtScale system begins to learn the behavior of the users and builds aggregates automatically. All the dimensions, filters, metrics, and hierarchies being used continuously feed the machine learning (ML) algorithms, and the custom-tailored aggregates get better as time goes on. This unique hybrid approach has a lot of advantages:
1. Accelerate Data-Driven Decisions at Scale
Users now have the ability to analyze data at the speed of thought; there’s no longer a process to wait days for cubes to be pre-built. Not only can users access all historical data, they can also continue to use their existing BI tools to achieve consistent query results and performance.
2. Control Complexity & Cost of Analytics
The ability to iterate quickly can make or break a business opportunity. With all the heavy lifting done by AtScale’s Autonomous Engineering instead of your valuable data engineers, your staff can pivot their work towards acquiring new data streams, strategize on new data services and providing advanced analytics. Let AtScale dynamically optimize the data rather than pin this manual effort on your data engineers. Improving the time to implement new dashboards and apply changes to existing ones is a nice bonus side effect.
3. One Consistent/Compliant View of Business Metrics & Definitions
It’s tempting to just provide tables and data and let the analysts mine the data, but there has to be a balance between data discovery and data delivery. What happens when you have conflicting answers to the same business question? Maybe it causes a delay in critical business initiatives, maybe it puts the entire company’s financial state at risk. Whatever the consequences may be, it’s best to avoid them by keeping business definitions managed in a single source of truth. AtScale acts as your single universal semantic layer that is platform independent; virtualize across all your data platforms while being able to point all kinds of SQL and MDX BI tools at it.
4. Mitigate Risks Associated with Data & Analytics
As data volume and variety grow, it’s natural for companies to have many disjointed systems storing data in silos. It puts an increasingly large burden on the IT team to secure the disparate systems as well as the data in transit. AtScale uses data virtualization to help stitch together all the sources while inheriting security frameworks from all underlying platforms. On top of providing complete lineage and audit trail of all consumption by users and queries, AtScale provides comprehensive role-based access controls and row/column/object level security.
Fast, consistent, secure, and maintenance-free analytics is no longer just a dream. The OLAP of the past has served us well, but it’s time for it to put on a robo-suit to keep up with the constantly evolving data world.
To learn more, Watch this on-demand webinar: Modernize Your Investment in SSAS Without Giving Up OLAP with AtScale Founder and CSO, Dave Mariani, as he explains more about OLAP and he demonstrates how AtScale Adaptive Analytics scales and modernizes your OLAP infrastructure.