OLAP is Back as Kyvos Insights

In 2013 I wrote a blog titled, Is OLAP Terminally Ill? I used the term “Terminally Ill” because I didn’t believe that the strategy of managing pre-aggregations was dead. Temporarily unnecessary, maybe, but not dead.

To be clear, by “OLAP” (Online Analytical Processing), I meant in both contexts of the software named SQL Server Analysis Services Multi-Dimensional (SSAS MD) and the activity of analyzing data with the expectation of sub-second query response times. The subject of OLAP being dead had been going around for a couple of years in the midst of the rise of in-memory technologies such as SAP HANA, “Hekaton”, and of course, Analysis Services Tabular (which debuted in SQL Server 2012). Additionally, Hadoop and Big Data were well along in becoming household words.

To summarize that blog, my primary argument was that OLAP wasn’t dead because there will always be a place for providing sub-second response time through pre-aggregation. It just so happened that the volume of data in 2013 was moving just beyond the easy reach of SSAS MD customers. That is beyond the capability of SSAS’s Cube Design Wizard, where life became very tough for the poor data engineers. The new technologies offered scalable (Hadoop) and/or potentially easier (Analysis Services Tabular) alternatives.

However, today in 2020, data volumes and variety are magnitudes crazier than in 2013. So pre-aggregation is making a comeback. Pre-aggregation promotes sub-second response time and preserves very expensive compute time by processing once and accessing many times.

I tried to leave the SSAS MD world when I wrote that blog 2013, but I kept finding contract work too good to pass up. The makings of a dinosaur. By 2016 I knew that I needed to fully move on to the Cloud. So I swore off SSAS MD and ended up working for customers with very large, on-prem, data warehouses ready to move to Azure. But there were always SSAS MD cubes somewhere in my customers’ enterprise that needed facelifts – so I would one again become the SSAS guy.

We would always discuss migration of the SSAS MD cubes to the Cloud. But to what? Azure Analysis Services (AAS) was the most obvious. In many instances, the cubes were simple and small enough to easily migrate to a reasonably-priced AAS resource.

Beyond AAS, there were a few roadblocks. Sometimes the customers were migrating to AWS or GCP and wished to be stick to one platform, as much as possible. Some customers were tired of the seemingly outdated notion of “cubes” having struggled with SSAS MD during the 2000s. Many customers that had been in the Cloud for a while became used to not having pre-existing limits on scalability. So much for AAS’s scale limits due to being in-memory.

Fast-forward to a few months ago, around August 2020. The Covid-19 lockdowns gave me time to think about highly scalable, Cloud-based alternatives to SSAS MD. I decided to write a comprehensive three-part blog on migrating from SSAS MD to some Cloud platform. I also intended on writing tools for the migration and I would offer guidance on several data warehouse platforms available on Azure; Synapse, Databricks, the various flavors of Azure SQL Server, of course, Snowflake, and even CosmosDB.

After writing much of the blog text on the situation and background for migration, it was time to research existing Cloud-based OLAP solutions. I searched for the key words, SSAS Migration OLAP Cloud, and right at the top was a recent blog titled, Kyvos Launches New Utility to Simplify Migration of SSAS Cubes to the Cloud. Oh!

I read through that migration document, watched a couple of videos on Kyvos, and looked at the leadership. It sure did look like a scale-out OLAP system, built for the Cloud, connections to many Cloud sources, a good level for native transformations, pre-aggregations, dimensions, measures, even MDX. Impressive. So much for my comprehensive migration blog and developing an SSAS Migration Tool.

I thought to contact the CEO, Praveen Kankariya, on LinkedIn to ask more about it and tell him my amusing story. As I began typing the message, I saw we had already communicated – back in 2014! I recalled that he did contact me after reading Is OLAP Terminally Ill? Long story short, a few weeks later (Sept 2020), I accepted a position at Kyvos Insights as a Principal Solution Architect (FTE).

Kyvos will feel rather familiar on the surface to SSAS MD folks. It’s not exactly an apples to apples comparison. Maybe gourds and pumpkins because pumpkins can grow to hundreds of pounds? After all, at the core, Kyvos is a scale-out platform versus SSAS MD being scale-up. Of course, scale-up means Kyvos cubes can theoretically grow to 100s of terabytes if needed.

There are some differences in functionality, interface, and particularly configuration. Kyvos is still, in the end, all about dimensions and measures and sub-second query responses. It’s actually easier to configure cubes than with SSAS. The power of MDX is there in all its glory. But for MDX haters, a SQL option is available, one that is more versatile than that of SSAS. Connection to your favorite BI tool is easy due to Kyvos’ use of XMLA and custom providers for tools such as Tableau and PowerBI.

I’ve been at Kyvos for two months now. I’m at the point where I feel I’ve dug deeply enough to be absolutely confident about the quality of Kyvos’ product, the immense value it can add to today’s data volumes and variety, and the incredibly dedicated and bright folks behind its development and implementation.

If you have any questions or would like to schedule a demo, please drop me an email at eugene.asahara@kyvos.io

Update – May 27, 2021

In the main post, I mention having written much of a three-part post on migration from SSAS MD to a few Cloud platforms. I never did finish it, and probably will not. However, I did write a three-part series on migration from SSAS MD to Kyvos:

• SSAS MD to Kyvos Migration – Part 1 – Cloud-Based OLAP Solution
• SSAS MD to Kyvos Migration – Part 2 – Overview of the Primary Objects
• SSAS MD to Kyvos Migration – Part 3 – Cube Design Experience