4.1 SAN Fabric Design, Zoning & Multi-Protocol Storage Networking

SAN Fabric Design, Zoning & Multi-Protocol Storage Networking

Storage Area Networks (SANs) form the mission-critical foundation of enterprise data centers, interconnecting host compute clusters with high-performance storage arrays. Unlike local area networks designed for general-purpose client-server communication, a SAN is purpose-built to deliver deterministic, low-latency, and lossless block I/O transport. Designing an enterprise SAN requires deep mastery of physical topologies, protocol layering, credit-based flow control, dynamic fabric addressing, granular zoning policies, and emerging protocols such as NVMe over Fabrics (NVMe-oF).

---

1. Fibre Channel Topologies: Architecture & Evolution

The Fibre Channel (FC) standard defines three distinct physical and logical interconnect topologies. Understanding their operational differences highlights why modern enterprise data centers standardize almost exclusively on switched fabrics.

1. Point-to-Point (FC-P2P)

• Architecture: Direct physical connection between exactly two Fibre Channel N_Ports (typically one host server HBA port connected directly to one storage array controller port via an optical fiber cable) without an intervening switch.
• Operational Characteristics: Provides the simplest configuration with zero switch management overhead and maximum available link bandwidth dedicated between the two nodes.
• Limitations: Highly restrictive scalability. Connecting N hosts to redundant storage controllers requires dedicated physical ports on the storage array for every host port, creating severe port exhaustion. FC-P2P is primarily used for direct array-to-array replication links or ultra-compact branch-office two-node clusters.

2. Arbitrated Loop (FC-AL)

• Architecture: Shared ring topology connecting up to 126 device ports (NL_Ports) and one switch fabric loop port (FL_Port) in a closed physical or logical loop.
• Addressing & Arbitration: Devices are identified using an 8-bit Arbitrated Loop Physical Address (AL_PA). When a device requires transmission, it must arbitrate for access to the shared medium; only one pair of devices can communicate across the loop at any given microsecond.
• Limitations: High latency under contention, complex arbitration delays, and vulnerability to catastrophic failure. A link break or malfunctioning transceiver anywhere on the loop disrupts the entire loop unless isolated by a managed hub with internal port bypass circuits. FC-AL is completely obsolete in modern SAN deployments and serves only as a legacy architectural reference.

3. Switched Fabric (FC-SW)

• Architecture: Interconnects hosts (N_Ports) and storage array controllers (N_Ports) through dedicated Fibre Channel switches using Fabric Ports (F_Ports). Multiple switches are linked together via Expansion Ports (E_Ports) across Inter-Switch Links (ISLs).
• Operational Characteristics: Every connected device enjoys dedicated, full-duplex bandwidth to the switch backplane. Devices communicate concurrently without loop arbitration. The topology scales to support up to 16 million theoretically addressable devices using dynamic 24-bit Fibre Channel IDs (FCIDs).
• Enterprise Adoption: FC-SW is the universal industry standard for enterprise SANs, delivering non-blocking line-rate performance, dynamic routing, and non-disruptive scalability.

---

2. Redundant Dual-Fabric Architecture (Fabric A & Fabric B Isolation)

High availability in SAN design is achieved by building two completely independent, concurrent, and isolated storage fabrics, universally designated as Fabric A and Fabric B.

+-------------------------------------------------------------------------+
|                 DUAL-FABRIC REDUNDANCY ARCHITECTURE                     |
+------------------------------------+------------------------------------+
| SAN FABRIC A (ISOLATED)            | SAN FABRIC B (ISOLATED)            |
+------------------------------------+------------------------------------+
| Switch A1 === [ISLs] === Switch A2 | Switch B1 === [ISLs] === Switch B2 |
| HBA Port 1 (Host Server)           | HBA Port 2 (Host Server)           |
| Controller Port A1 (Storage Array) | Controller Port A2 (Storage Array) |
| Controller Port B1 (Storage Array) | Controller Port B2 (Storage Array) |
+------------------------------------+------------------------------------+
| ABSOLUTE RULE: ZERO INTER-SWITCH LINKS (ISLs) BETWEEN FABRIC A & B      |
+-------------------------------------------------------------------------+

The Cardinal Rule of SAN Design: Physical Air-Gap Isolation

• Zero Cross-Fabric ISLs: Fabric A and Fabric B must never be physically linked with an Inter-Switch Link (ISL). Bridging the fabrics merges them into a single logical Fibre Channel fabric sharing a single Principal Switch, a common Name Server database, and a single fabric configuration.
• Fault Domain Containment: If an ISL were placed between Fabric A and Fabric B, a catastrophic event—such as a rogue switch configuration overwrite, a corrupted zone set push, a looping fabric condition, or a Registered State Change Notification (RSCN) storm—would instantly propagate across both fabrics simultaneously, causing complete storage outage across the entire data center.
• Symmetrical Redundancy: Every host server must install a dual-port Host Bus Adapter (or two single-port HBAs across independent PCIe root complexes). Port 1 connects strictly to Fabric A; Port 2 connects strictly to Fabric B. Similarly, enterprise storage controllers present front-end target ports symmetrically across both fabrics.
• Host-Based Multipathing (MPIO): Storage high availability is managed exclusively at the host operating system level via multipathing software (such as Microsoft MPIO, VMware Native Multipathing Plugin [NMP], or Linux device-mapper-multipath). The host detects redundant physical paths across both isolated fabrics to the same underlying LUN, using Asymmetric Logical Unit Access (ALUA) to balance I/O across Active/Optimized paths while maintaining instantaneous failover to Active/Non-Optimized paths.

---

3. Fibre Channel Layered Architecture & Flow Control

Fibre Channel is structured into five distinct protocol layers, ranging from physical transceivers to upper-layer protocol mappings:

+-------------------------------------------------------------------------+
|                     FIBRE CHANNEL PROTOCOL LAYERS                       |
+--------+--------------------------+-------------------------------------+
| LAYER  | NAME                     | PRIMARY FUNCTION                    |
+--------+--------------------------+-------------------------------------+
| FC-4   | Upper Layer Protocol Map | SCSI-3 (FCP), NVMe (FC-NVMe)        |
| FC-3   | Common Services          | Striping, encryption, multicasting  |
| FC-2   | Framing & Signaling      | Frame format, BB_Credits, flow ctrl |
| FC-1   | Transmission Protocol    | 8b/10b, 64b/66b line encoding       |
| FC-0   | Physical Interface       | Cables (OM3/OM4/OS2), SFP optics    |
+--------+--------------------------+-------------------------------------+

FC-0: Physical Interface

Defines physical media, optical transceivers (SFP+, SFP28 for 32G, SFP56 for 64G), and transmission distances. Multi-mode optical fiber (OM3, OM4, OM5) utilizing 850 nm Short Wavelength (SW) lasers supports distances up to 100-150 meters within data centers. Single-mode optical fiber (OS2) utilizing 1310 nm Long Wavelength (LW) lasers supports campus and metropolitan distances up to 10-40 kilometers.

FC-1: Transmission Protocol

Defines serial data transmission, bit synchronization, word alignment, and line encoding schemes. Legacy speeds (1G, 2G, 4G, 8G) employ 8b/10b encoding (introducing 20% framing overhead). Modern speeds use far more efficient encodings: 16G and 32G use 64b/66b (about 3% overhead), and 64G uses 256b/257b with PAM4 signaling, paired with Forward Error Correction (FEC) to recover corrupted bits in real time.

FC-2: Framing & Signaling (The Heart of Fibre Channel)

Defines the core 2148-byte Fibre Channel frame format, consisting of a 4-byte Start of Frame (SOF) delimiter, a 24-byte Frame Header, a variable Data Payload (0 to 2112 bytes), a 4-byte Cyclic Redundancy Check (CRC), and a 4-byte End of Frame (EOF) delimiter.

Flow Control via Buffer-to-Buffer Credits (BB_Credits)

Standard Ethernet relies on dropping packets when intermediate switch buffers fill up, forcing higher-layer protocols like TCP to detect loss and retransmit. Fibre Channel eliminates frame loss due to buffer congestion entirely through a deterministic, hardware-level flow control mechanism called Buffer-to-Buffer Credits (BB_Credits):

1. Credit Negotiation: During port initialization and Fabric Login (FLOGI), two directly connected ports (e.g., host HBA port and switch port) negotiate the number of receive buffers available at each endpoint. If the switch port has 16 receive buffers, it grants 16 BB_Credits to the host transmitter.
2. Credit Transmission: Every time the host transmits a Fibre Channel frame, it decrements its local BB_Credit counter by 1. If the credit counter reaches 0, the transmitter pauses transmission immediately.
3. Credit Replenishment: When the switch receiver successfully processes a frame and flushes it from its receive buffer into the switch fabric, it transmits a 4-byte primitive signal called Receiver Ready (R_RDY) back to the host.
4. Credit Restoration: Upon receiving the R_RDY primitive, the host transmitter increments its credit counter by 1 and resumes transmission. Frames are never dropped due to full buffers.

The Slow-Drain Device Challenge & Credit Exhaustion

While BB_Credits guarantee lossless transport, they introduce the risk of slow-drain devices. If an edge device (such as an overloaded server HBA or degraded optical transceiver) cannot process frames at line rate, its receive buffers remain full. The device stops returning R_RDY primitives to the switch. As a result:

• The switch port's BB_Credit counter depletes to zero, halting frame egress.
• Incoming frames destined for that slow port accumulate in the switch's internal ingress buffers.
• When internal switch buffers fill, the switch must withhold R_RDY primitives across upstream ISLs to adjacent switches.
• Congestion spreads backward across the entire fabric, stalling I/O to completely healthy, unrelated servers and arrays. Modern SAN switches mitigate this through automated fabric monitoring (e.g., Brocade Fabric Vision / Cisco Congestion Avoidance), which detects credit stall conditions, alerts administrators, throttles the offending port, or isolates the link via port fencing.

FC-3: Common Services

Provides common services across multiple ports, including link striping (aggregating multiple physical links into a single logical channel), multi-cast transmission, and broadcast primitives.

FC-4: Upper Layer Protocol (ULP) Mapping

Maps enterprise application protocols into standard Fibre Channel frames. The primary FC-4 protocols are FCP (Fibre Channel Protocol), which encapsulates SCSI-3 Command Descriptor Blocks (CDBs), and FC-NVMe, which encapsulates low-latency NVMe command queues.

---

4. Fibre Channel Addressing & Fabric Login Workflow

Fibre Channel implements a dual-tier addressing structure: permanent hardware identifiers for unique physical identification, and dynamic logical identifiers for high-speed hardware frame routing.

WWNN vs. WWPN

• World Wide Node Name (WWNN): A globally unique 64-bit hexadecimal identifier (e.g., 20:00:00:10:9B:12:34:56) assigned by the manufacturer to an entire physical device or chassis. A dual-port HBA or multi-controller storage array possesses a single WWNN representing the hardware unit.
• World Wide Port Name (WWPN): A globally unique 64-bit hexadecimal identifier assigned to each individual physical or virtual port interface on an HBA or storage controller (e.g., Port 1 is 10:00:00:10:9B:12:34:56, Port 2 is 10:00:00:10:9B:12:34:57). Fibre Channel zoning and LUN masking configurations are built using WWPNs, never WWNNs.

The 24-Bit Fibre Channel ID (FCID)

Because routing frames using 64-bit WWPNs would require massive routing tables and excessive processing latency in switch ASICs, the fabric dynamically assigns a compact 24-bit Fibre Channel ID (FCID) (also known as the S_ID / D_ID) to every device during fabric login. The 24-bit FCID consists of three 8-bit octets:

+-----------------------+-----------------------+-----------------------+
| DOMAIN_ID (Bits 23-16)| AREA_ID (Bits 15-8)   | PORT_ID (Bits 7-0)    |
+-----------------------+-----------------------+-----------------------+
| Switch Identifier     | Port Group / ASIC     | Individual Interface  |
| (e.g., 0x01 to 0xEF)  | (e.g., 0x00 to 0xFF)  | (e.g., 0x00 to 0xFF)  |
+-----------------------+-----------------------+-----------------------+

1. Domain_ID (Bits 23 to 16): Uniquely identifies the specific switch within the fabric. Valid Domain IDs range from 1 to 239 (0x01 to 0xEF). Values 0 and 240-255 are reserved for fabric services.
2. Area_ID (Bits 15 to 8): Identifies a group of ports, a line card, or a specific switch ASIC within the chassis.
3. Port_ID (Bits 7 to 0): Identifies the individual physical or virtual port interface within that area.

The Three-Step Fabric Login Workflow

Before a host initiator can communicate with a storage target, it must execute a sequential three-step login process:

1. FLOGI (Fabric Login):
   • The host N_Port transmits an FLOGI frame to the well-known fabric controller address 0xFFFFF7.
   • The switch and host exchange operational parameters (such as supported classes of service and BB_Credits).
   • The switch assigns a unique 24-bit FCID to the host port based on the local switch Domain_ID and port location.
2. PLOGI (Port Login):
   • The host initiates a Port Login to the well-known Name Server address 0xFFFFFC.
   • The host queries the fabric Name Server to discover available target devices matching its assigned zone.
   • The host then issues a direct PLOGI frame to the specific target port's FCID to establish an N_Port-to-N_Port communication session, negotiating maximum payload size and sequence parameters.
3. PRLI (Process Login):
   • Establishes the upper-layer protocol (FC-4) operational context between initiator and target.
   • For traditional block storage, PRLI activates FCP-SCSI; for modern flash arrays, it negotiates FC-NVMe command processing and queue parameters.

---

5. Fabric Zoning Architecture & Enterprise Best Practices

Fabric zoning is a hardware-enforced access control service configured on Fibre Channel switches that partitions the fabric into logical, isolated communication groups.

The Critical Function of Zoning: RSCN Containment

Beyond preventing unauthorized server-to-storage access, zoning performs a vital operational role: Registered State Change Notification (RSCN) containment.

• When any SAN device logs in, disconnects, reboots, or suffers a link flap, the switch fabric controller generates an RSCN frame to inform endpoints that the Name Server database has changed.
• In an unzoned fabric, every single device receives the RSCN broadcast. Upon receiving an RSCN, every host HBA and storage controller halts ongoing queue processing to re-query the Name Server and rediscover all targets. In an enterprise SAN with hundreds of servers, a single flapping cable can cause an RSCN storm, triggering rolling I/O timeouts and application crashes.
• Fabric zoning isolates RSCNs: switch hardware delivers RSCN frames only to devices that reside within the same active zone as the flapping port, completely shielding the rest of the data center.

Zoning Classification: Soft Zoning vs. Hard Zoning

+-------------------------------------------------------------------------+
|                         ZONING CLASSIFICATIONS                          |
+------------------------------------+------------------------------------+
| SOFT ZONING (WWPN-BASED)           | HARD ZONING (PORT-BASED)           |
+------------------------------------+------------------------------------+
| Members defined by 64-bit WWPNs.   | Members defined by physical switch |
| Enterprise Best Practice.          | ports (Domain, Index/Port).        |
| Survives cable moves to other      | Moving a cable breaks connectivity |
| ports on the switch fabric.        | instantly.                         |
| Enforced by switch Name Server.    | Enforced by hardware switch ASIC.  |
+------------------------------------+------------------------------------+

Zoning Architectures: Single-Initiator vs. Multi-Initiator

1. Single-Initiator / Single-Target (SI-ST) Zoning (The Golden Standard)

• Rule: Each zone contains exactly one host initiator WWPN and one storage array target WWPN.
• Benefits: Guarantees absolute fault isolation. Initiators can never communicate with other initiators, eliminating initiator-to-initiator cross-talk, bus reset interference, and Name Server churn. If an array port experiences link degradation, only the specific zone containing that port is affected.
• Drawback: Increases the total number of zones requiring administrative creation (though modern automation tools and scripting eliminate manual overhead).

2. Single-Initiator / Multiple-Target (SI-MT) Zoning (Widely Accepted)

• Rule: Each zone contains one host initiator WWPN and multiple storage array target WWPNs (e.g., Target Port 1 on Controller A and Target Port 1 on Controller B).
• Benefits: Reduces total zone count while preventing initiator-to-initiator cross-talk.
• Drawback: An RSCN triggered by one array port is delivered to all targets in that zone.

3. Multiple-Initiator / Multiple-Target (MI-MT) Zoning (Strict Anti-Pattern)

• Rule: Placing multiple host initiators and multiple storage targets into a single zone.
• Verdict: Strictly prohibited in enterprise production SANs. Hosts constantly probe each other, waste fabric bandwidth, and propagate RSCN storms across all compute nodes.

4. Smart Zoning / Peer Zoning (Modern Optimization)

• Developed by switch vendors (Brocade Peer Zoning, Cisco Smart Zoning) to combine the administrative simplicity of large zones with the hardware isolation of SI-ST zoning.
• Administrators configure a single zone containing multiple initiators and targets, but designate each member as an Initiator or a Target/Principal.
• The switch ASIC automatically programs the internal routing hardware to permit communication only between Initiators and Targets, completely blocking Initiator-to-Initiator traffic and suppressing cross-initiator RSCNs.

---

6. SAN Security: Fabric Zoning vs. Storage Controller LUN Masking

A foundational concept on technical storage certification exams is understanding that SAN security requires two distinct, non-overlapping enforcement tiers:

+-------------------------------------------------------------------------+
|                        TWO-TIER SAN ACCESS CONTROL                      |
+------------------------------------+------------------------------------+
| TIER 1: FABRIC ZONING              | TIER 2: LUN MASKING                |
+------------------------------------+------------------------------------+
| Location: Fibre Channel Switches   | Location: Storage Array Controller |
| Scope: Fabric Port Visibility      | Scope: Volume / LUN Presentation   |
| Mechanism: WWPN / FCID Routing     | Mechanism: WWPN / IQN Authorization|
| Purpose: Restricts which nodes can | Purpose: Authorizes which specific |
| discover and target each other;    | logical volumes a host can mount;  |
| isolates fabric RSCNs.             | prevents unshared data corruption. |
+------------------------------------+------------------------------------+

• Why Zoning Alone Is Insufficient: Zoning only controls whether Host A can see Controller Port 1. If Controller Port 1 presents 50 different LUNs belonging to 50 different departments, zoning cannot prevent Host A from seeing and corrupting all 50 LUNs.
• Why LUN Masking Alone Is Insufficient: LUN masking runs entirely inside the array controller software. Without switch zoning, all 500 servers in a data center would log into the array target ports simultaneously, exhausting the storage controller's login tables, flooding the fabric with Name Server traffic, and exposing the array to RSCN storms.

---

7. NVMe over Fabrics (NVMe-oF): FC-NVMe and NVMe/TCP

NVMe over Fabrics (NVMe-oF) extends the ultra-low latency and extreme parallelism of local NVMe SSDs across storage network fabrics, replacing the legacy SCSI-3 command architecture.

+-------------------------------------------------------------------------+
|                       SCSI-3 vs. NVMe ARCHITECTURE                      |
+------------------------------------+------------------------------------+
| LEGACY SCSI-3 PROTOCOL (FCP/iSCSI) | MODERN NVMe-oF PROTOCOL            |
+------------------------------------+------------------------------------+
| Single Command Queue per device    | Up to 64,000 Parallel Queues       |
| Queue Depth: 32 to 128 commands    | Queue Depth: 64,000 commands/queue |
| Global lock contention across CPUs | Lockless per-CPU-core queue mapping|
| High protocol translation overhead | Direct submission/completion rings |
+------------------------------------+------------------------------------+

1. FC-NVMe (NVMe over Fibre Channel)

• Architecture: Encapsulates NVMe command and completion frames directly inside FC-4 frames, operating concurrently alongside traditional SCSI FCP traffic across the exact same 32Gb/64Gb physical Fibre Channel infrastructure.
• Benefits: Delivers sub-millisecond flash latency and massive IOPS without requiring new switch hardware. Leverages existing Fibre Channel BB_Credit lossless delivery, fabric zoning, and enterprise reliability.
• Requirements: Requires modern Gen6 (32Gb) or Gen7 (64Gb) HBAs (e.g., HPE SN1610E) and switch firmware supporting FC-NVMe.

2. NVMe/TCP (NVMe over Standard Ethernet TCP/IP)

• Architecture: Transports NVMe command queues encapsulated directly inside standard TCP/IP packets (default port TCP 8009 or 4420) over standard 25GbE, 100GbE, or 200GbE Ethernet switches.
• Benefits: Completely eliminates the need for expensive proprietary Fibre Channel HBAs and dedicated optical switches. Delivers dramatically lower latency and CPU overhead than iSCSI by eliminating SCSI translation.
• Trade-offs: Unlike FC-NVMe, NVMe/TCP operates over lossy standard Ethernet unless paired with Data Center Bridging (DCB) enhancements such as Priority-based Flow Control (PFC) and Explicit Congestion Notification (ECN).

---

8. iSCSI Storage Networking: Addressing, Discovery & Best Practices

Internet Small Computer Systems Interface (iSCSI) encapsulates SCSI-3 Command Descriptor Blocks (CDBs) within standard TCP/IP packets over listening port TCP 3260, enabling block SAN deployment across commodity Ethernet hardware.

iSCSI Addressing: IQN and EUI Formats

• iSCSI Qualified Name (IQN): The standard ASCII naming convention (up to 255 characters) composed of four distinct fields:
  • iqn.yyyy-mm.reverse-domain-name:optional-unique-string
  • Example: iqn.2026-10.com.example:storage:alletra-mp-01
  • iqn: Protocol identifier.
  • 2026-10: The year and month when the naming authority registered the domain.
  • com.example: Reverse domain name of the organization.
  • :storage:alletra-mp-01: Storage administrator's unique target or initiator naming string.
• Extended Unique Identifier (EUI): An alternative 16-character hexadecimal format based on IEEE EUI-64 (e.g., eui.02004567A42541BC).

Target Discovery Mechanisms

• SendTargets: The host initiator connects to a known target portal (IP address and TCP port 3260) and issues an iSCSI text command requesting available storage targets. The portal responds with a list of available IQNs and target interfaces.
• Internet Storage Name Service (iSNS): A centralized directory server protocol (operating over TCP/UDP port 3205) that dynamically registers iSCSI initiators and targets. When an initiator boots, it queries the iSNS server to automatically discover its authorized targets, mirroring the automated discovery behavior of Fibre Channel Name Servers.

Enterprise iSCSI Engineering Best Practices

1. Network Isolation: iSCSI traffic must never share broadcast domains with standard corporate LAN traffic. Dedicated physical switches or dedicated storage VLANs are strongly recommended to isolate storage traffic from broadcast storms and user-traffic congestion.
2. Jumbo Frames (MTU 9000): Standard Ethernet frames use an MTU of 1500 bytes. Because storage payloads transfer large 4KB, 8KB, or 64KB block chunks, standard frames split every transfer into many more packets, raising per-packet processing and interrupt overhead. Enabling Jumbo Frames (MTU 9000) end-to-end (host NIC, virtual switch, physical switch trunk ports, and storage target ports) carries six times as much payload per frame, cuts the number of packets by roughly 6x, raises wire efficiency from about 95% to about 99%, and reduces host CPU utilization. Critical Warning: If MTU 9000 is not configured consistently across every hop, packets are dropped, leading to silent connection failure.
3. Data Center Bridging (DCB): Deploying DCB switches with Priority-based Flow Control (PFC - IEEE 802.1Qbb) provides hardware-level pause frames for specific VLAN priority tags (Class of Service), delivering lossless Ethernet transport similar to Fibre Channel BB_Credits.
4. Authentication & Security: iSCSI operates without inherent link encryption. Deployments enforce CHAP (Challenge Handshake Authentication Protocol):
   • One-Way (Target) CHAP: Target authenticates the host initiator before granting access.
   • Mutual (Bi-Directional) CHAP: Host and target mutually authenticate each other using independent pre-shared secrets, preventing rogue array spoofing on the IP network.

---

9. Comprehensive SAN Protocol Comparison Matrix

---

10. Real-World SAN Deployment Scenario: Mission-Critical Database Cluster

Architectural Challenge

A financial services firm is deploying a four-node Oracle RAC database cluster that requires deterministic sub-millisecond response times, zero data loss, and uninterrupted operation during hardware maintenance. The storage architect must design the SAN fabric connectivity to an enterprise all-flash storage array.

Implemented Architecture & Design Decisions

1. Redundant Fabric Topology: Two physically separate Fibre Channel fabrics (Fabric A and Fabric B) are constructed using dual HPE B-series 32Gb switches. No Inter-Switch Links connect Fabric A to Fabric B.
2. Host Connectivity: Each database server is equipped with two dual-port 32Gb FC HBAs. Port 1 of HBA 1 and Port 1 of HBA 2 connect to Fabric A switches; Port 2 of HBA 1 and Port 2 of HBA 2 connect to Fabric B switches, providing bus-level PCIe redundancy.
3. Zoning Policy: Single-Initiator / Single-Target (SI-ST) soft zoning is implemented across both fabrics using host and array WWPNs. Each server port resides in a dedicated zone with exactly one storage controller target port. This configuration limits the RSCN broadcast domain and completely isolates server-to-server traffic.
4. Storage Presentation: Storage controllers configure LUN masking to bind the shared database cluster volume (LUN 10) strictly to the eight authorized server WWPNs across both fabrics, ensuring no other enterprise hosts can access the raw database extents.
5. Multipathing Configuration: Host multipathing is configured with ALUA Active/Optimized paths distributed across both Fabric A and Fabric B controllers, providing seamless sub-second path failover if a switch port, transceiver, or storage controller fails.

---

11. Common Exam Traps & Architectural Pitfalls

• Exam Trap 1: Attempting to Interconnect Fabric A and Fabric B with an ISL. Exam questions often present scenarios where an engineer connects an ISL between Fabric A and Fabric B to "share bandwidth" or "provide switch failover." This is a catastrophic failure mode. It eliminates physical isolation, merges fabrics into a single failure domain, triggers zone database synchronization conflicts, and allows an RSCN storm or configuration error to take down both fabrics.
• Exam Trap 2: Believing LUN Masking Replaces Fabric Zoning. Candidates frequently assume that configuring LUN masking on the storage array eliminates the need for switch zoning. LUN masking enforces volume access rights, but without fabric zoning, all host HBAs freely discover array target ports, register with the Name Server, exhaust array target login buffers, and receive fabric-wide RSCN broadcasts during link events.
• Exam Trap 3: Confusing WWNN and WWPN in Zoning Configurations. Fabric zoning configurations must always specify the World Wide Port Name (WWPN). A World Wide Node Name (WWNN) represents the physical enclosure or host chassis and is never routable or zonable on Fibre Channel switches.
• Exam Trap 4: Assuming Jumbo Frames (MTU 9000) Can Be Configured Only on the Storage Array. In an iSCSI environment, configuring MTU 9000 on the storage array while leaving host NICs or intermediate switch ports at standard MTU 1500 causes massive packet fragmentation and packet drops. Jumbo Frames must be enabled consistently across every single component in the communication path (host OS, hypervisor vSwitch, physical switch ports/trunks, and storage controllers).