Introduction

Oracle Real Application Clusters (RAC) is one of the most powerful high-availability database technologies available today. It allows multiple database instances to run on different nodes while accessing the same database storage.

But behind this seamless experience lies a complex internal architecture that ensures consistency, synchronization, and performance across nodes.

In this article, we’ll break down three critical Oracle RAC internals in a simple and practical way:

• GRD (Global Resource Directory)
• GES (Global Enqueue Service)
• GCS (Global Cache Service)

By the end, you’ll understand how Oracle RAC manages concurrency and ensures data consistency across all instances.

---

What Problem Does RAC Solve?

In a single-instance database, one database manages all data access. But in RAC:

👉 Multiple instances access the same data blocks simultaneously

This creates challenges like:

• Data consistency
• Lock coordination
• Cache synchronization
• Performance overhead

Oracle solves this using GRD, GES, and GCS

---

What is GRD (Global Resource Directory)?

The Global Resource Directory (GRD) is the central brain of RAC coordination.

It keeps track of:

• Who owns which data block
• Which instance holds locks
• Which instance has cached copies
• Which instance is the master of a resource

Think of GRD like this:

GRD = Cluster-wide dictionary of all resources

Every RAC instance maintains a portion of the GRD, so it is:

✔ Distributed
✔ Highly available
✔ Efficient

---

What is GES (Global Enqueue Service)?

GES is responsible for global locks across RAC instances.

It manages:

• Row locks
• Table locks
• DDL locks
• Library cache locks

Whenever two instances try to modify the same data or structure:

👉 GES coordinates who gets the lock

Example

Instance 1 updates a row
Instance 2 tries to update same row

GES ensures:

Only one instance holds the lock at a time

This prevents data corruption and conflicts

---

What is GCS (Global Cache Service)?

GCS handles data block consistency across instances

This is also known as Cache Fusion

Instead of writing blocks to disk, RAC transfers them directly between nodes over the interconnect.

Example

Instance 1 has a block in memory
Instance 2 needs it

👉 GCS transfers the block from Instance 1 → Instance 2

No disk read required 🔥

---

GRD + GES + GCS Relationship

Here’s how everything works together:

GRD → knows where everything isGES → manages locksGCS → manages data blocks

---

Where are these components stored?

Oracle uses different memory areas:

Component	Stored in
GES (Enqueue Resources)	Shared Pool
GCS (Cache Resources)	Large Pool
GRD	Distributed across instances

---

Master and Shadow Resources

Each resource in GRD has:

• One MASTER copy
• Multiple SHADOW copies

Example

Block A:Master → Instance 2
Shadow → Instance 1, Instance 3

This allows fast coordination without central bottlenecks.

---

How RAC Handles a Block Update

Let’s walk through a simple scenario:

Step 1: Instance 1 updates a block

• GCS marks it as modified
• GRD updates ownership

Step 2: Instance 2 requests same block

• GRD identifies Instance 1 as owner
• GCS transfers block via interconnect

Step 3: Lock managed by GES

• GES ensures safe access

Result:

✔ No disk read
✔ Fast transfer
✔ Consistent data

---

Why This Matters for Performance

These components directly impact:

• Query performance
• Transaction speed
• Interconnect traffic
• Scalability

Poor design causes:

• Excessive block pinging
• High global cache waits
• Enqueue contention

Good design ensures:

• Cache locality
• Reduced cross-node traffic
• High throughput

---

Common Performance Waits Related to GRD

When something is not optimized, you may see waits like:

• gc cr request
• gc buffer busy
• enq: TX row lock contention
• ges resource busy

These are directly related to GCS and GES activity

---

Best Practices for RAC Performance

Here are a few DBA tips:

✔ Use service-based workload routing

Keep related sessions on the same instance

✔ Avoid hot blocks

Partition heavily accessed tables

✔ Use NOORDER sequences in RAC

Reduces SQ enqueue contention

✔ Increase sequence cache size

Reduces global synchronization

✔ Monitor interconnect traffic

Ensure low latency network

---

RAC Internals Simplified

Let’s summarize everything in one simple model:

Client → connects to InstanceInstance uses:
   GES → to get locks
   GCS → to get data blocks
   GRD → to find resource owner

---

Easy Memory Trick

Remember this:

GES = Locks
GCS = Blocks
GRD = Directory

---

Conclusion

Oracle RAC is powerful because it allows multiple instances to work together on the same database.

But this is only possible because of the intelligent coordination done by:

• GRD (Global Resource Directory)
• GES (Global Enqueue Service)
• GCS (Global Cache Service)

These components ensure:

✔ Data consistency
✔ High performance
✔ Scalability
✔ Fault tolerance

Understanding these internals will help you:

• Troubleshoot RAC issues faster
• Improve performance
• Design better architectures