General questions
What is Oracle Database 23ai?
Oracle Database 23ai is the next long-term support release of Oracle Database, superseding 19c. This release brings AI to your data with the addition of AI Vector Search to Oracle’s converged database. This capability combined with new unified development paradigms and mission-critical capabilities makes it simple for developers and data professionals to power apps, application development, and mission-critical workloads with AI.
What are the key features of Oracle Database 23ai?
Oracle Database 23ai brings more than 300 new features compared to the previous generation. Here are some of the key features. You can find links to more details about all these features in this blog post on Oracle Database 23ai: Where to find information:
- AI Vector Search: Oracle AI Vector Search is a new capability introduced in Oracle Database 23ai. AI Vector Search is a collection of features that includes a new vector data type, vector indexes, and vector search SQL operators that enable Oracle Database to store the semantic content of documents, images, and other unstructured data as vectors and use these to run fast similarity queries. These new capabilities also support retrieval augmented generation (RAG), a breakthrough generative AI technique that combines large language models (LLMs) and private business data to deliver responses to natural language questions Learn more about AI Vector Search
- JSON Relational Duality Views: JSON Relational Duality Views unifies the relational and document data models to provide the best of both worlds. Developers can operate on the same underlying data as either JSON documents, using standard document APIs, or as relational, using standard SQL. Learn more about JSON Relational Duality View
- Operational property graphs: Oracle Database offers native support for property graph data structures and graph queries. Developers can now easily build graphs and graph applications in conjunction with transactional data, JSON, Spatial, and other data types supported in the database. In addition, the new ANSI standard SQL/PGQ lets you build and query these graph relationships in an easy-to-understand language.
- SQL Firewall: Oracle Database offers native support for property graph data structures and graph queries. Developers can now easily build graphs and graph applications in conjunction with transactional data, JSON, Spatial, and other data types supported in the database. In addition, the new ANSI standard SQL/PGQ lets you build and query these graph relationships in an easy-to-understand language. Learn more about SQL Firewall
- True Cache: This solution simplifies caching against Oracle Database. Developers simply deploy a diskless True Cache instance to their midtier and let Oracle Database manage the rest. Using Active Data Guard (ADG) technology, Oracle Database 23ai determines whether the cache is consistent with the latest changes in the primary instance and retrieves data if it’s not present in the cache. Learn more about True Cache
- SQL enhancements: Oracle Database 23ai includes new features such as Application Usage Domains, which allow developers to define what columns represent, such as emails, credit cards, zip codes, and more. And schema annotations provide developers with a means to centralize rich and descriptive metadata for their objects, a new Boolean data type, direct joins in updates, and much more.
- Oracle RAC scalability and availability enhancements: Oracle Database 23ai introduces enhancements that further solidify Oracle's commitment to delivering unparalleled availability and resilience for mission-critical databases.
- High availability
- Local Rolling Database Maintenance supports uninterrupted database availability during maintenance activities (such as patching) for Oracle Real Application Clusters (RAC) and Oracle RAC One Node databases. This significantly improves the availability of databases while limiting the impact on other nodes in the cluster.
- Oracle RAC Two-Stage Rolling Updates allow nonrolling patches to be applied after patching all nodes, thereby reducing application downtime.
- Globally Distributed Database with Raft Replication: Enables rapid failover within seconds and zero data loss during node or data center outages, facilitating an Active-Active-Active symmetric distributed database architecture that enhances availability, simplifies management, and optimizes resource utilization globally.
- Scalability
- Oracle RAC Ordered Sequences Optimization improves the performance of workloads that use ordered sequences.
- Oracle RAC Smart Connection Rebalancing measures and redirects sessions accessing the same database objects to the instance where the object is cached, thereby reducing inter-instance traffic, and significantly improving the performance of high-contention workloads.
Can you explain AI Vector Search in more detail?
AI Vector Search uses vectors to enable fast and simple queries on business data in conjunction with LLMs. To better understand this functionality, check out the following sub-questions. See more details about AI Vector Search
- What is a vector?
A vector is a numerical representation of text, images, audio, or video that encodes the meaning of the data, not the underlying words or pixels.
- Can vectors be created inside the database?
Yes, via the vector_embedding() SQL function.
Choose Sentence-Transformer embedding models from Hugging Face and then securely upload them to the database.
- Can vectors be created outside the database?
Yes, using both commercial and open source models with either REST calls or local libraries.
Embedding models that have been tested include openai.com, cohere.com, Hugging Face Transformers, Sentence-Transformers, Transformers.js, and using the ONNX Runtime.
- Are GPUs needed to create vectors?
Vectors can be created using CPUs or GPUs.
- Which LLMs does AI Vector Search work with?
AI Vector Search should be able to work with any LLM.
OpenAI GPT 3.5 and 4.0, Cohere, Llama2, Mistral, Gemini, Palm2, and Vertex AI LLMs have all been tested.
Some of these LLMs are available through Oracle Cloud Infrastructure (OCI) Generative AI service.
- Which embedding models does AI Vector Search work with?
More than 90 embedding models from OpenAI, Cohere, Transformer, Sentence Transformer, Transformer.js, Xenova, FastEmbed, and the ONNX Runtime have been tested.
- Does AI Vector Search support LangChain or LlamaIindex?
Oracle AI Vector Search supports the Oracle AI Vector Search provider for LangChain and LlamaIndex.
What other products are being announced with Oracle Database 23ai?
This announcement focuses on cloud first, developer first.
OCI:
- Oracle Database 23ai on Exadata Cloud@Customer with Exadata Systems Software 24ai
- Oracle Database 23ai on Exadata Database Service with Exadata Systems Software 24ai
- Oracle Database 23ai on Base Database Service
- OCI GoldenGate 23ai
Azure:
- Oracle Database 23ai on Exadata Database Service with Exadata Systems Software 24ai
Developers:
- Always Free Autonomous Database23ai
- Autonomous Database 23ai Free Container Image
- Oracle Database 23ai Free
- GoldenGate 23ai Free
What is the difference between Oracle Database 23c announced last year and Oracle Database 23ai?
Given the importance of the breakthrough AI technology in this release, we are renaming Database 23c as Database 23ai. Oracle Database 23ai is the version 23.4.0.0.0 and will be referred to as such moving forward.
Why should you upgrade from Oracle Database 19c to 23ai?
Oracle Database 23ai is the next long-term support release of Oracle Database, superseding 19c. A long-term release of Oracle Database offers five years of premium support and three years of extended support. Oracle Database 19c Premier Support ends on April 30, 2024, and while Oracle waived the additional fees for Extended Support from May 1, 2024 to April 30, 2026, Extended Support will end on April 30, 2027, giving you a longer support overlap with Oracle Database 23ai.
Are there any upgrade considerations from a previous version to Oracle Database 23ai?
Oracle provides upgrade paths from either Oracle Database 19c or Oracle Database 21c to 23ai. All other versions of the database need to upgrade to one of the versions listed prior, and then to 23ai. The specific upgrade process may vary depending on the current version and configuration of your particular Oracle Database deployment.
Is there a free trial available for Oracle Database 23ai?
Yes, Oracle offers free trials for Oracle Database 23ai through Oracle Cloud Infrastructure (OCI) with Autonomous Database Free Tier. Users can sign up for a free trial account and explore the features and capabilities of Oracle Database 23ai on OCI. Alternatively, Oracle Database 23ai Free is available for download as a Linux RPM file, an Oracle Linux-based Docker image, Oracle Virtual Box VM, or packed with Vagrant.