Duplexing of Coupling Facility (CF) structures in IBM Parallel Sysplex is a high-availability technique that keeps two synchronized copies of a CF structure so that the system can continue operating even if one CF or structure instance fails.

It is essentially:

A real-time replication mechanism for CF structures across two Coupling Facilities (or two instances) to eliminate single points of failure.

1. What is being duplexed?

A CF structure can be:

• Lock structure
• Cache structure
• List structure

With duplexing enabled:

• One structure is the primary
• The other is the secondary (backup copy)

Both exist in separate CFs or separate CF logical instances.

2. Why duplexing is needed

Without duplexing:

• CF failure = loss of structure state
• leads to recovery delays or service interruption

With duplexing:

• No loss of state
• Instant failover capability

👉 Goal: continuous availability of sysplex services

3. Types of duplexing

A. Synchronous duplexing (full duplex)

• Every update is written to both CF structures at the same time
• Both copies stay identical in real time

✔ Highest consistency
❌ Slightly higher latency

B. Asynchronous duplexing (less common conceptually)

• Primary CF is updated first
• Secondary is updated shortly after

✔ Lower latency
❌ Small window of inconsistency (rarely used for critical structures)

4. How duplexing works (step-by-step)

Step 1: Request arrives

A z/OS system issues a CF operation:

• lock request
• cache update
• list update

Step 2: Primary CF processes request

• Primary CF executes the operation immediately
• Updates its structure state

Step 3: Secondary CF replication

• Same operation is sent to secondary CF
• Secondary applies identical update

Step 4: Acknowledgment

• Only after both CFs confirm (in synchronous mode)
• Request completes back to system

5. What is actually kept in sync?

Depending on structure type:

Lock structures

• Lock ownership state
• Wait queues
• Lock compatibility matrix

Cache structures

• Data blocks
• Directory entries (ownership/sharing state)
• Valid/invalid flags

List structures

• Queue entries
• Order of elements
• Head/tail pointers

6. Failure handling (key benefit)

If primary CF fails:

• Secondary CF becomes active immediately
• All systems switch transparently
• No loss of structure state

👉 This is called CF structure recovery by takeover

7. Duplexing modes in practice

IBM Z supports:

A. User-managed duplexing

• z/OS application or sysplex services control it

B. System-managed duplexing

• CF automatically maintains synchronization
• Preferred in modern systems

8. Performance trade-offs

9. Why duplexing is different from normal replication

Unlike database replication:

• CF duplexing is synchronous and hardware-assisted
• Happens at microsecond scale
• Maintains exact structural state, not just data

👉 It is more like mirrored shared memory, not log-based replication.

10. Relationship to sysplex resilience

Duplexing works alongside:

• XCF (communication + membership)
• STP (time consistency)
• CF lock/cache structures (shared state)

Together they ensure:

• no single point of failure
• continuous data availability
• seamless workload continuity

11. Simple mental model

Think of CF duplexing as:

A real-time mirrored memory system where every structural update is instantly copied to a second Coupling Facility, allowing instant failover with zero loss of shared state.

Key takeaway

Duplexing of CF structures ensures:

• Real-time replication of CF lock/cache/list structures
• Synchronous consistency between two Coupling Facilities
• Instant failover without losing shared system state
• High availability for Parallel Sysplex workloads