racle Database Licensing: The Complete Guide

punitoracledba.blogspot.com / Oracle · Exadata · ExaCC · Licensing · Performance

Deep Dive

Oracle Database Licensing:
The Complete Guide
Every DBA Needs to Read

From the two core metrics through Exadata OCPUs, VMware traps, cloud nuances, and what really happens in an LMS audit.

7 Parts 4 Architecture Diagrams punitoracledba

Table of Contents

1. The Two Metrics — NUP vs Processor
2. The Core Factor Table
3. Editions — What You're Actually Buying
4. Exadata & ExaCC Architecture
5. OCPUs and vCPUs Explained
6. Rules Everyone Gets Wrong
7. Cloud, Audits & Staying Compliant

If you've spent any time in the Oracle ecosystem, you already know the feeling. Someone asks how many licenses the company needs, and suddenly the room gets very quiet. Oracle licensing is one of those topics that makes experienced DBAs nervous, finance teams confused, and Oracle's License Management Services (LMS) team very, very happy.

This article is my attempt to demystify it completely — with full architecture diagrams showing how each piece connects. We'll go from the absolute basics all the way through Exadata and ExaCC licensing nuances, the traps that catch even experienced teams, and what happens when Oracle knocks on your door for an audit.

Part 1

The Two Metrics — How Oracle Counts What You Owe

Everything in Oracle licensing starts with one question: how do you measure usage? Oracle gives you two options, and picking the wrong one can cost you significantly.

Named User Plus (NUP)

A Named User Plus license covers one specific person — or one device — authorized to access the Oracle Database. It doesn't matter whether that person logs in every day or once a year. If they could access it, they need a license. One person, one license, regardless of how many connections or sessions they open.

Minimum Rule

There is a minimum of 25 NUP licenses per Processor license you would otherwise need. If your server requires 8 Processor licenses, you must buy at least 200 NUP licenses — even if you only have 15 actual users.

Processor

Processor licensing ignores users entirely. You pay based on the number of CPU cores that Oracle software is running on — or, more precisely, that Oracle software could run on. For most enterprise systems — anything with a web tier, an API layer, hundreds of application users — Processor licensing is the only practical choice.

# Core formula for Processor licensing Licenses needed = physical cores × core factor # Full cost formula Year 1 cost = (licenses × list price) + (licenses × list price × 22%) Annual after = licenses × list price × 22% # support only

Part 2

The Core Factor Table — Not All Cores Are Equal

Oracle publishes the Oracle Processor Core Factor Table, assigning a multiplier to each CPU architecture. The most important row is Intel/AMD — which is what virtually every modern server runs.

CPU Architecture	Core Factor	Example
Intel / AMD (x86-64)	0.5	Exadata, ExaCC, most cloud
Oracle SPARC T-series	0.25 – 0.5	SPARC T8 servers
Oracle SPARC S/M/High-end	0.5	SPARC M8
IBM POWER	1.0	AIX servers
HP PA-RISC	1.0	HP-UX servers

# Example: 2-socket Intel server, 16 cores per socket Physical cores = 2 sockets × 16 cores = 32 cores Core factor = 0.5 (Intel x86-64) Processor licenses needed = 32 × 0.5 = 16 licenses # Exadata / ExaCC with OCPUs (Intel) 1 OCPU = 1 physical Intel core = 0.5 Processor licenses 8 OCPUs enabled → 8 × 0.5 = 4 Processor licenses

Part 3

Editions — What You're Actually Buying

Standard Edition 2 (SE2)

SE2 is Oracle's entry-level offering with hard limits: maximum 2 populated sockets, no Oracle RAC (as of 19c), and no advanced options. Licensed per Named User Plus or per socket — not per core.

Enterprise Edition (EE)

EE is the full product — no socket limits, full RAC support, full feature availability. But here's the critical thing: buying EE does not mean you've licensed all the features. Many features you think are "just part of Oracle" are separate licensed options sold on top of EE.

Enterprise Edition Options & Management Packs (each sold separately)

Option / Pack	Type	Common Use
Real Application Clusters (RAC)	Option	HA / scale-out
Active Data Guard	Option	Standby reads + switchover
Advanced Compression	Option	Table / index compression
Partitioning	Option	Range/list/hash partitions
Advanced Security	Option	TDE, network encryption
In-Memory	Option	Columnar in-memory store
Diagnostics Pack	Pack	AWR, ASH, ADDM
Tuning Pack	Pack	SQL Tuning Advisor
Data Masking Pack	Pack	Mask sensitive data

Part 4

Exadata & ExaCC Architecture

To understand Oracle licensing in the context of engineered systems, you need to see what Exadata and ExaCC actually look like under the hood. The diagrams below show both — and why their OCPU model simplifies licensing compared to running Oracle on VMware.

Diagram 1 — Oracle Exadata Database Machine (full rack)

Diagram 2 — Oracle Exadata Cloud@Customer (ExaCC)

The key licensing insight from these diagrams: Exadata and ExaCC use hard partitioning at the hardware level. OCPUs you don't enable are not licensed. You pay only for what you turn on. This eliminates the VMware soft-partitioning trap entirely.

Part 5

OCPUs and vCPUs — Why the Numbers Look Different

One of the most common points of confusion when moving to OCI or ExaCC is the OCPU count. Someone from AWS says "we have 32 vCPUs" and expects to need 32 OCPUs. They actually need 16. Here's why.

Diagram 3 — OCPU vs vCPU: what each unit really represents

This matters enormously for licensing calculations. If you migrate from AWS and say "we were on 32 vCPUs," you only need 16 OCPUs on ExaCC to match the same physical core count. And at 0.5 factor, that's only 8 Processor licenses.

Part 6

The Rules Everyone Gets Wrong

Rule 1 — Soft Partitioning: The Most Expensive Misunderstanding in Enterprise IT

Many DBAs assume: "I have a VMware VM with 4 vCPUs assigned. Oracle only runs in that VM. So I only need to license 4 vCPUs." Oracle disagrees. Strongly.

VMware is classified as soft partitioning — because the VM can theoretically migrate to any host in the cluster via vMotion. Oracle's position: you must license every physical core on every host in the cluster the VM could ever migrate to.

Real Cost Shock

vSphere cluster: 10 hosts × 32 cores × 0.5 factor = 160 Processor licenses required — even if your Oracle VM only uses 4 vCPUs. At ~$25K list price per EE Processor license, that's $4M in licenses for what looks like a small Oracle VM.

Hard partitioning technologies Oracle accepts: Oracle VM (OVM) with hard-partitioned domains, Solaris Zones, IBM LPAR (specific configs), physical servers, and Exadata/ExaCC (by design).

Rule 2 — The Enterprise Manager Diagnostics Pack Trap

This catches DBAs constantly. When you connect to OEM Cloud Control, you'll see beautiful dashboards — AWR reports, ASH analytics, the SQL Tuning Advisor. Every one of those requires either the Diagnostics Pack or Tuning Pack license.

Audit Evidence

Oracle's LMS team will pull dba_feature_usage_statistics from your database. This view records every licensed feature accessed — with timestamps. Two years of Monday AWR reports without a Diagnostics Pack = two years of back-licensing + support fees.

Rule 3 — Support Is Not Optional

Oracle charges 22% of the license cost per year for Software Update License & Support. Stop paying and you lose access to security patches. Rejoining means paying all missed years at full rate before Oracle reinstates you.

Rule 4 — Dev and Test Environments Need Licenses Too

Oracle requires you to license all environments where Oracle software is installed and running — including dev, test, QA, and staging — unless you use the OTN Developer License (which is restricted to a single developer machine and prohibits production-scale data).

Part 7

Cloud, Audits & Staying Compliant

Diagram 4 — Oracle licensing: complete one-frame map

Cloud Licensing — OCI, ExaCC, AWS, Azure

Platform	Model	License included?	Options extra?
OCI (BYOL)	Use existing on-prem licenses	No — you bring	Yes
OCI (License Included)	Bundled hourly rate	Yes	Yes
ExaCC	OCPU-based, EE bundled	EE included	RAC, ADG, etc. extra
AWS (BYOL)	Authorized cloud environment	No — you bring	Yes
AWS (License Included)	RDS / higher instance cost	Yes	Limited options
Azure (BYOL)	Authorized cloud environment	No — you bring	Yes

Oracle LMS Audits — What Actually Happens

Oracle can initiate an audit at any time with 45 days' notice — this is in your license agreement. Audits are often triggered after renewal negotiations, a large purchase, or simply as part of Oracle's scheduled programme. They can also arrive disguised as a "Customer Experience Review."

The LMS team will ask you to run Oracle-provided scripts that collect hardware inventory, software versions, dba_feature_usage_statistics, listener configurations, and VM configuration data. They cross-reference all of this against your license entitlement.

Most common audit findings:

• Unlicensed Diagnostics/Tuning Pack usage (most common by far)
• VMware soft-partitioning gap (often the largest dollar value)
• Unlicensed Options — especially Partitioning and Advanced Compression
• Under-counted NUP users
• Test/dev environments not licensed

Staying Compliant — Practical Checklist

• 1
  Run Oracle's lms_collect.sql (from My Oracle Support) against your estate quarterly. Know your position before Oracle does.
• 2
  Configure OEM licensing settings to disable Diagnostics Pack and Tuning Pack features if you haven't licensed them. This prevents accidental usage that creates audit exposure.
• 3
  Classify every server: hard partitioned (Exadata, physical, OVM) vs soft (VMware, Hyper-V). Document it. This is your first line of defence in any audit.
• 4
  Track all licenses in a CMDB — CSI numbers, effective dates, edition, options. Oracle's My Oracle Support license portal should match what you track internally.
• 5
  Negotiate aggressively at renewal time. Oracle's list prices are rarely what anyone pays. Renewal is your leverage point — use it. Consider an independent Oracle licensing specialist before any large negotiation.
• 6
  If an audit notice arrives, engage an independent Oracle licensing specialist before you respond. The way you respond to LMS — and what you share — matters enormously.

ExaCC Advantage

ExaCC sidesteps the VMware problem entirely (hard partitioning at hardware level), bundles EE into the OCPU price, and gives you OCI metering for predictable monthly billing. For organizations that need on-prem data residency with Oracle workloads, it's worth the total cost of ownership comparison.

Wrapping Up

Oracle licensing is deliberately complex — that complexity creates audit leverage and upsell opportunities from Oracle's perspective. But once you understand the model (metric × core factor × edition × options × support), you can navigate it.

# The mental model to hold Processor licenses = physical cores × 0.5 (Intel/AMD) ExaCC: 1 OCPU = 1 core = 0.5 Processor licenses (EE bundled) VMware: license the entire cluster, not just the VM AWR in OEM without Diagnostics Pack = compliance violation Support = 22%/year of license cost, non-optional in practice

Stay licensed, stay informed — and always read the fine print before you click anything in Enterprise Manager.

punitoracledba.blogspot.com  ·  Oracle · Exadata · ExaCC · Performance · Licensing

Vector Search in Oracle Database 26ai – Learning Concepts for DBAs and Enterprise Teams

A practical learning article for Oracle DBAs, architects, and anyone starting the Oracle 26ai journey

---

Introduction

In the AI era, the way we search data is changing rapidly. Traditional search methods depend heavily on exact words, exact patterns, or strict SQL conditions. But modern applications need something smarter. They need systems that can understand meaning, context, and similarity.

That is where Vector Search becomes one of the most important innovations in Oracle Database 26ai. For DBAs, this is not just a new feature. It is a new learning area that connects database technology with AI-driven applications.

Simple idea: Vector Search helps the database search by meaning, not only by exact keywords.

---

What is Vector Search?

Vector Search is a method of finding information based on similarity. Instead of asking, “Does this row contain the exact text?”, the database asks, “Which rows are most similar in meaning to this query?”

In simple terms, text, images, documents, and other content can be converted into mathematical representations called vectors. These vectors capture the meaning or characteristics of the data. Once stored, the database can compare them and return the closest matches.

Traditional Search Example

Query: payment issue
Result: Only rows containing the exact same or very similar words

Vector Search Example

Query: payment issue
Result: payment failed, invoice error, billing problem, transaction declined

This is the power of semantic understanding.

---

Why Vector Search Matters in Oracle 26ai

Oracle Database 26ai is moving toward intelligent data platforms. Vector Search is one of the core concepts behind that transformation. It is important because enterprise systems now need to support:

• AI assistants and chatbots
• Semantic search across documents and knowledge bases
• Recommendation engines
• Fraud pattern analysis
• Retrieval-Augmented Generation (RAG) applications
• Smarter support systems

Without vector search, these use cases usually require external search engines or AI platforms. With Oracle 26ai, the database itself becomes part of the intelligent search layer.

Learning point: Vector Search is not replacing SQL. It is adding a new intelligent search capability to the database.

---

Core Learning Concepts You Must Understand

1. What is a Vector?

A vector is a numerical representation of data. It can represent text, images, audio, or any content in a mathematical form. In AI systems, vectors are often used to represent meaning and context.

For example, two sentences with similar meaning may have vectors that are close to each other, even if the words are different.

Example:
“Customer payment failed” and “Transaction was declined” may be stored as different text,
but their vectors may be close because their meaning is related.

2. What are Embeddings?

Embeddings are the vector representations created from data. When text or content is processed through an AI model, that model generates an embedding which captures the meaning of the content.

These embeddings are then stored in the database and used for similarity search.

3. What is Similarity Search?

Similarity Search means comparing one vector with others and finding which ones are closest. The closer the vectors, the more similar the meaning or characteristics.

This is very different from traditional filtering using equals, like, or regular expressions.

4. What is a Vector Index?

A vector index is a specialized structure designed to speed up similarity searches. Just like a normal database index helps with fast row retrieval, a vector index helps the database quickly find the nearest matching vectors.

For large datasets, vector indexing becomes extremely important for performance.

5. What is Semantic Search?

Semantic Search means searching by meaning instead of exact words. It is one of the biggest use cases for vector technology. This helps enterprise applications return more relevant results for users.

---

How Vector Search Works – Step by Step

Step 1: Data is collected (documents, tickets, notes, logs, product descriptions)

Step 2: AI model converts that data into embeddings (vectors)

Step 3: Vectors are stored in the database

Step 4: A vector index is created for faster search

Step 5: User query is also converted into a vector

Step 6: Database compares vectors and returns the nearest results

This flow allows Oracle Database 26ai to support intelligent retrieval directly from the database layer.

---

Simple Architecture View

Traditional Search Model

User Query → SQL LIKE / Exact Match → Limited Results

Oracle 26ai Vector Search Model

User Query → Convert to Vector → Compare with Stored Embeddings → Similarity Search → Intelligent Results

---

Conceptual SQL Example

Below is a simple conceptual example to understand how vector-based storage and search may look.

Step 1: Create a Table

CREATE TABLE support_tickets (
    ticket_id NUMBER,
    description CLOB,
    embedding VECTOR
);
  

Step 2: Insert Sample Data

INSERT INTO support_tickets VALUES (
    1,
    'Payment failed during checkout',
    VECTOR_EMBEDDING('Payment failed during checkout')
);

INSERT INTO support_tickets VALUES (
    2,
    'Invoice generation error for customer',
    VECTOR_EMBEDDING('Invoice generation error for customer')
);
  

Step 3: Create a Vector Index

CREATE VECTOR INDEX idx_support_embedding
ON support_tickets(embedding);
  

Step 4: Run a Similarity Search

SELECT ticket_id, description
FROM support_tickets
ORDER BY VECTOR_DISTANCE(
    embedding,
    VECTOR_EMBEDDING('payment issue')
)
FETCH FIRST 5 ROWS ONLY;
  

Important note: The SQL above is for learning concepts and simple understanding. Exact syntax and implementation may vary depending on Oracle feature usage and environment design.

---

Where DBAs Will See Real Value

From a DBA perspective, vector search becomes valuable when business teams want more intelligent applications without building a completely separate AI search platform.

Practical Use Cases

Support Ticket Search
Find similar historical incidents even when wording is different

Enterprise Knowledge Search
Search internal documents, SOPs, and runbooks intelligently

Fraud and Risk Analysis
Detect similar suspicious patterns faster

Product Recommendation
Suggest related products or services based on similarity

In enterprise environments, this can significantly improve decision-making, user experience, and search relevance.

---

What DBAs Need to Learn in This Area

As Oracle Database 26ai evolves, DBAs do not need to become data scientists overnight. But they should understand the core concepts well enough to support AI-enabled platforms.

• Basic understanding of embeddings and vectors
• How vector indexes affect storage and performance
• How AI-driven applications may query the database
• Security and governance considerations for AI workloads
• How to test, monitor, and validate performance

DBA mindset shift: The role is moving from only managing data storage to supporting intelligent data retrieval.

---

Performance and Design Considerations

Like any new feature, vector search must be evaluated carefully before production rollout.

• Vector indexes may consume significant storage
• Memory and CPU impact must be tested
• Large-scale datasets require careful design
• Search relevance should be validated with business users
• Security and access controls remain important

DBAs should treat vector search just like any important new database capability: learn it, test it, tune it, and implement it carefully.

---

Learning Summary

• Vector Search means searching by similarity and meaning
• Vectors are numerical representations of content
• Embeddings are vectors generated from text or other data
• Vector indexes improve performance for similarity search
• Semantic Search helps applications return more relevant results
• Oracle 26ai brings this capability into the database world

---

My Perspective

From an Oracle DBA and enterprise operations perspective, Vector Search is one of the most exciting learning areas in Oracle Database 26ai. It is not just a technical feature. It represents how the database is becoming more intelligent and more connected to modern AI workloads.

For DBAs, learning this concept early is a strong investment. It helps us understand where enterprise database architecture is heading next.

Final message: If you understand Vector Search, you understand one of the core building blocks of the AI database era.

---

Conclusion

Oracle Database 26ai is introducing capabilities that take the database far beyond traditional storage and query processing. Vector Search is one of those major steps. It enables intelligent retrieval, semantic matching, and support for next-generation AI applications.

For Oracle professionals, this is the right time to learn the concept deeply, understand the terminology, and prepare for a future where databases do much more than store rows and columns.

---

Suggested Labels / Tags

Oracle Database 26ai: Transforming Databases into Intelligent Platforms

A complete guide for DBAs, Architects, and Enterprise Leaders

---

🔹 Introduction

The release of Oracle Database 26ai marks a major shift in how databases are used in modern enterprises. This is no longer just a database—it is an AI-powered intelligent data platform.

Key Idea: Databases are evolving from data storage systems to decision-making engines.

---

What Makes Oracle 26ai Special?

• AI inside the database
• Vector-based semantic search
• Advanced automation
• Reduced architecture complexity

---

Architecture Evolution

Traditional Architecture

Application → Database → ETL → AI Tools → Reports

Oracle 26ai Architecture

Application → Oracle 26ai (DB + AI + Vector Search) → Insights

Fewer components, faster performance, better security.

---

Core Features

Vector Search (Game Changer)

Vector search allows the database to understand meaning instead of exact keywords.

Example:
Search: "payment issue"
Results: billing error, invoice problem, failed transaction

• Semantic search
• AI chatbot support
• RAG applications

---

2) AI Vector Indexes

• Optimized for similarity search
• Handles large datasets
• High-performance AI queries

---

3) AI Built Inside Database

No need to move data outside. AI processing happens within the database.

• Better security
• Faster insights
• Simplified architecture

---

4️) Intelligent Automation

• Auto query tuning
• Predictive diagnostics
• Performance optimization

Impact: Less manual work for DBAs

---

DBA Perspective

Before

• Backup & Recovery
• Patching
• Performance tuning

After 26ai

• AI workload management
• Vector index understanding
• AI-enabled architecture support

Transformation: DBA → AI-Aware Data Engineer

---

Enterprise Use Cases

Customer Support
Semantic search across tickets

Fraud Detection
Pattern recognition

Enterprise Search
Smart document retrieval

Predictive Analytics
Future insights

---

Adoption Considerations

• Compatibility with existing systems
• Learning curve
• Infrastructure readiness
• Clear business use case

---

💡 My Perspective

After working extensively on Oracle EBS, DR setups, and enterprise databases, I see Oracle 26ai as the biggest shift after Multitenant architecture.

This is not optional learning anymore.
It is the future of the DBA role.

---

Final Thoughts

Oracle Database 26ai transforms the database into:

• ✔ Intelligent
• ✔ AI-driven
• ✔ Decision-making platform

Conclusion: The database is no longer just storing data—it is understanding it.

---

Tags

How Oracle's 26AI release transforms DBA workflows with in-database AI inference, vector search, autonomous tuning, and GenAI integration

/* ═════════════════════ ARCHITECTURE DIAGRAM ════════════════════════*/

Oracle 26AI —
The AI-Native Database Era

How Oracle's 26AI release transforms DBA workflows with in-database AI inference, vector search, autonomous tuning, and GenAI integration.

What is Oracle 26AI?

Oracle 26AI (also referred to as Oracle Database 26c with AI Extensions) is the most ambitious Oracle release since Autonomous Database. It embeds a full AI/ML inference engine directly inside the kernel — meaning your SQL queries can now invoke foundation models, vector similarity search, and AI-driven query rewriting without leaving the database tier.

💡 DBA Insight: For the first time, you don't need a separate Python microservice or REST call to run an LLM. DBMS_AI.GENERATE() is a native PL/SQL package — callable from SQL*Plus, APEX, EBS, or any OCI connection.

Key Pillars of Oracle 26AI

🧠

In-Database AI Inference

Run ONNX-format ML models directly inside Oracle using DBMS_AI. Zero data movement, zero latency penalty.

🔍

Native Vector Search

New VECTOR datatype with HNSW and IVF index types. Power RAG pipelines without leaving SQL.

⚙️

AI-Assisted SQL Tuning

AWR data now feeds an AI model that auto-rewrites suboptimal SQL and predicts execution plan regressions.

🔒

Security AI

Anomaly detection on session behaviour integrated with Oracle DB Security and Vault — self-healing policies.

🌩️

OCI GenAI Bridge

Call Cohere, Meta LLaMA, or OpenAI-compatible endpoints via DBMS_AI.GENERATE() from PL/SQL.

🛡️

Data Guard AI Failover

ML-based redo lag prediction triggers proactive switchover before service impact reaches end users.

New SQL & PL/SQL Syntax

1. Create a Vector Column

-- New VECTOR datatype (26AI)
CREATE TABLE product_embeddings (
  id          NUMBER GENERATED ALWAYS AS IDENTITY,
  product_id  NUMBER,
  description CLOB,
  embed       VECTOR(1536, FLOAT32)   -- 1536-dim OpenAI embedding
);

-- HNSW vector index
CREATE VECTOR INDEX idx_embed
  ON product_embeddings(embed)
  USING HNSW
  WITH TARGET ACCURACY 95;

2. Semantic Similarity Search

-- Find top-5 products similar to a query embedding
SELECT product_id, description,
       VECTOR_DISTANCE(embed, :query_vector, COSINE) AS score
FROM   product_embeddings
ORDER  BY score
FETCH FIRST 5 ROWS ONLY;

3. Call a Foundation Model from PL/SQL

DECLARE
  v_prompt   VARCHAR2(4000) := 'Summarise this AWR report in 3 bullet points: ' || :awr_text;
  v_response CLOB;
BEGIN
  v_response := DBMS_AI.GENERATE(
    provider   => 'OCI_GENAI',
    model      => 'cohere.command-r-plus',
    prompt     => v_prompt,
    max_tokens => 512
  );
  DBMS_OUTPUT.PUT_LINE(v_response);
END;
/

DBA Impact: What Changes for You

Area	Before 26AI	With 26AI
ML Model Scoring	Export to Python / R microservice	DBMS_AI.SCORE() inside SQL
Vector / Semantic Search	External pgvector or Pinecone	Native VECTOR type + indexes
SQL Tuning	Manual AWR analysis + hints	AI Tuning Advisor auto-rewrites SQL
Anomaly Detection	Custom SIEM integration	Built-in Security AI, zero config
GenAI Calls	App-tier REST with data copying	DBMS_AI.GENERATE() in PL/SQL
Data Guard Failover	Threshold-based DBA rules	ML lag prediction → proactive switch

Upgrade Path from 19c / 21c

Oracle 26AI supports in-place upgrade via DBUA from 19c (19.24+) and 21c. The AI extensions are licensed separately under the Oracle AI Database Services option — check with your Oracle rep before enabling DBMS_AI in production.

# Pre-upgrade compatibility check
./runInstaller -silent -checkSysPrereqs \
  -paramFile /u01/app/oracle/product/26.0.0/dbhome_1/assistants/dbua/dbua.rsp

# Enable AI extensions post-upgrade (CDB level)
sqlplus / as sysdba
ALTER SYSTEM SET ai_enabled = TRUE SCOPE=SPFILE;
SHUTDOWN IMMEDIATE;
STARTUP;

⚠️ Production Warning: Always test AI features on a non-production clone first. The AI Tuning Advisor's auto-rewrite feature is opt-in and should be validated via SQL Performance Analyser (SPA) before enabling on critical OLTP workloads.

Summary

Oracle 26AI is not a marketing rebrand — it is a genuine architectural shift. The database tier is becoming the AI execution layer, collapsing the boundary between data storage and intelligence. For DBAs, this means new packages to master (DBMS_AI, DBMS_VECTOR), new index types to manage (HNSW, IVF), and new responsibilities around AI governance and model lifecycle inside the DB.

The good news: if you already know Data Guard, AWR, and OML, you are 70% of the way there. The 26AI additions are evolutionary, not revolutionary — Oracle kept the DBA at the centre.

— punitoracledba  |  Database Architect · OCI · Exadata · EBS · Never Stop Sharing, Learning and Growing