6 months ago
64
At Oracle Cloudworld 2024, AI agents stole the show as the next big thing, with over 50 new AI agents unveiled across functions. Not just that. Oracle’s ambition of moving towards complete autonomous databases became clearer than ever.
“We can do a better job of protecting our data if the database system managing that data is fully autonomous. This is something that Oracle has been working on for a very long time and we think we’re quite good at it,” said Oracle chief Larry Ellison at the event.
Tirthankar Lahiri, senior vice president of mission-critical data and AI engines, spoke to AIM on the sidelines of the event in Las Vegas. Elaborating on the progress of Oracle’s autonomous database systems, he said, “I think without autonomous systems, it will be impossible to operate the data management system for the future.” Lahiri believes this to be critical to the evolution of the way businesses manage their data.
Oracle vs the World
Autonomous database systems are not a new concept for the organisation. First launched in 2017 at the Oracle Cloudworld event, the database has been employed by many large-scale companies, like Siemens, Reuters, Lyft, and Vodafone Fiji, across domains.
When asked how Oracle positions itself in the database market against strong contenders such as MongoDB and Redis, Lahiri mentioned that their approach doesn’t require compromise in data storage.
“We’re not logged into one-size-fits-all because of the database choices in how they store it. We think the way you store data should be the most optimal always, and independent of the way you access it. And, the one that lets you support multiple workload types on the same database,” said Lahiri.
He said document databases, such as MongoDB, create ‘a lot of data duplication’ by storing data in a repetitive manner. In contrast, Oracle’s system avoids these inefficiencies by using unique, normalised tables while still allowing developers the flexibility of document-style access through JSON views.
“So, Oracle’s big advantage, I think, is this converged platform that lets you support multiple APIs, multiple data personalities, and multiple developer styles,” he said.
He also emphasised on the scalability of its platform, which can handle different workloads and departments without requiring multiple databases unless necessary for security or isolation.
RAG and LLMs in Autonomous Systems
“LLMs, when paired with autonomous databases, provide a seamless way for users to interact with data using natural language, eliminating the need for technical expertise,” said Lahiri when discussing the relevance of LLMs and RAGsmin autonomous databases where query handling, data analysis, and decision-making can be executed.
While optimistic, Lahiri also brought up the limitations of RAG in autonomous database systems. “We do use built-in machine learning algorithms to find in the database, but we really don’t see RAG as a vehicle for improving database performance, per se, because it will just take too long,” he said.
Acknowledging LLM capabilities, such as excelling at interactive communication, and real-time decision-making for database tuning and optimisation, relies on built-in machine learning. Lahiri confirmed that though they don’t use LLMs for these tasks, the reverse is true—where their databases enhance RAG capabilities.
Database Trajectory at Oracle
Tirthankar Lahiri in a screengrab from a video interview with AIM at CloudWorld 2024.
An IIT-Kharagpur graduate and PhD holder from Stanford University in the 90s, Lahiri has been with Oracle for close to three decades and has witnessed massive technological changes in the tech and cloud industry. While critical nuances such as security scalability and others in database management remain the same, a lot of other things have emerged.
“What has changed, I think, is that the role of the database has just broadened over the past three decades,” he said.
Lahiri also observed that when he joined Oracle, the focus was on foundational features like bitmap indexes and shared disk architecture, which allowed applications to scale without rewriting code, which pretty much laid the groundwork for Oracle’s success in handling large-scale applications.
Oracle is leveraging its existing robust technologies, like query optimisers and data formats, to power AI-driven features such as vector search. “So that’s what I see as a change. Try not to go back and redo the fundamentals, but reuse them in new and creative ways to solve the next generation of problems,” he said.
Lahiri also highlighted India as a “huge market” for Oracle, attributing this to rapid industrial digitisation and the increasing need for data management. He praised the country’s adoption of UPI, noting its impact on Oracle’s innovations, including AI-based transaction prioritisation for managing vast payment volumes.
