Revolutionizing Database Migrations: Insights from the Datometry OpenDB team at AWS re:Invent 2023

Authors: Gourab Mitra and Jozsef Patvarczki 

AWS re:Invent 2023 was yet again an exhilarating experience. A flurry of new offerings was announced during the conference, such as adding vector capabilities to existing products, serverless offerings, and, of course, Generative AI. 

However, we didn’t attend to gawk at shiny new objects. We were there to get a download on database migrations. Yes, the discipline of moving from one database to another. We wanted to talk to practitioners in the trenches to understand the issues they are running into. We were eager to learn the gaps in the products and processes they use for migration. After all, migrations are a space that is both business-critical and excruciating at once. 

Looking back, we can summarize our experience at AWS re:Invent 2023 as three major observations. 

Moving away from Oracle remains an unsolved problem 

If somebody walked into the conference without knowing what it was about, they might have thought they were at an Oracle event. Oracle databases and the complexity that comes with them were omnipresent. Entire sessions were dedicated to Oracle migrations in varying levels of detail.  

It became clear during almost every Q&A session that customers are extremely hesitant about changing their Oracle installations. Most attendees professed a desire to move away from Oracle—but would settle for re-platforming their Oracle to RDS for now. In effect, they are trading their existing Oracle for a fully managed version. 

Oracle’s licensing, known for its complexity and nuance, is to blame here. Customers spend considerable time devising licensing strategies. Unfortunately, customers get so confused and distracted by this tinkering that they take their eyes off the prize. Instead of looking to move away from Oracle altogether, they optimize their pain. 

Static code converters have plateaued

Static code converters that translate queries written in a legacy SQL dialect to the SQL of a modern system seemed all the rage in the past decade. In theory, Oracle queries can be translated to, say, PostgreSQL queries this way. AWS’s contributions to this discipline are Data Migration Service and Schema Conversion Tool. 

First, the good news. The static code converters presented at the conference boast impressive coverage. Depending on the workload, they may achieve up to 80%. AWS has clearly advanced the state of the art. Coverage of 80% is probably close to the theoretical maximum for static code conversion. However, this number is considerably lower when dynamic SQL or complex PL/SQL expressions are present. 

On the downside, 80% is still a non-starter when you consider that database migrations are a classic 80-20-problem. The easy, or initial, 80% of a conversion may be significant in volume but represent only a fraction of the required overall effort. The remaining 20% of the migration problem is where 80% of the effort goes. It looks like code conversion has entered an era of diminishing returns. 

Most mission-critical workloads are still on-premises

Enterprises are moving their workloads to the cloud. After all, that made the cloud business one of the biggest growth markets in history. But upon closer inspection, there is one curious wrinkle. Ask any IT leader about their mission-critical workloads, and chances are they will admit they still run those on-premises. 

They already moved various workloads to the cloud, but the mission-critical database workloads are still the exception. Re-platforming to a cloud-hosted database is not enough of a benefit unless one migrates to a true cloud-native database. And therein lies the problem. Moving from on-prem Oracle to hosted Oracle is an okay move. Moving from Oracle to, say, PostgreSQL is a true game changer. 

OpenDB: A Pragmatic Approach 

AWS re:Invent 2023 made it quite clear that database migrations are still one of the biggest obstacles in modern IT. The high cost and the enormous risk of migrations hold enterprises back and mean new vendors are missing out on revenue opportunities. 

For us at Datometry, this was once more confirmation that solving the database migration problem is a priority for every enterprise. With OpenDB, we’ve launched the first multi-modal database that can run both Oracle and PostgreSQL workloads. Enterprises can move from Oracle to OpenDB without changing their application code. 

The future of database migrations is efficient, affordable, and seamlessly managed. Datometry OpenDB is now available on AWS marketplace.