Cisco Nexus 5600 10-Gbps Switches 40GbE, FCoE and 1µs Latency

The Cisco Nexus 5600 Platform 10-Gbps switches are data center access and aggregation switches for Top-of-Rack, Middle-of-Row, and End-of-Row deployments requiring 10-Gbps Ethernet, 40-Gbps uplinks, Layer 2 and Layer 3 forwarding, FCoE, VXLAN, FabricPath, and Cisco Fabric Extender integration.

The platform consists of 1RU and 2RU chassis designs. It is intended for enterprise data centers, service-provider hosting, cloud infrastructure, virtualization, high-performance computing, high-frequency trading, and converged LAN and SAN environments. The switches use Cisco NX-OS and provide line-rate forwarding, cut-through latency of approximately 1 microsecond, shared packet buffering, network programmability, and hardware-based Layer 2 and Layer 3 services.

Product selection

The Cisco Nexus 5672UP and Cisco Nexus 5672UP-16G support up to 72 10-Gbps interfaces when 40-Gbps QSFP+ ports are configured with breakout cables. The 56128P supports 40-Gbps fixed ports and can add further 10-Gbps, Fibre Channel, and 40-Gbps interfaces through the N56-M24UP2Q expansion module.

All three chassis support Ethernet and FCoE on their Ethernet-capable ports. The unified ports provide native Fibre Channel according to the port and model restrictions shown in the interface matrix.

Hardware and interface details

Cisco Nexus 5672UP

The 5672UP provides 48 fixed 1/10-Gbps SFP+ ports. The last 16 ports are unified ports supporting 2-, 4-, and 8-Gbps Fibre Channel, 10-Gbps Ethernet, and FCoE. All 48 fixed ports support classical Ethernet and FCoE. Six fixed QSFP+ ports support 40-Gbps Ethernet and FCoE and can be converted to 10-Gbps interfaces using QSFP+ breakout cables.

The switch has two 1+1 redundant, hot-swappable power supplies and three hot-swappable independent fan modules with 2+1 redundancy. Port-side intake and fan-side intake airflow options are available.

The platform provides up to 15 buffer-to-buffer credits per port and supports SAN extension up to 3.7 kilometers at 8-Gbps Fibre Channel speed, assuming full Fibre Channel frames of 2112 bytes.

Cisco Nexus 5672UP-16G

The 5672UP-16G provides 48 fixed 10-Gbps SFP+ ports. The first 24 ports support 1/10-Gbps Ethernet. The last 24 ports are unified and support 2-, 4-, 8-, and 16-Gbps Fibre Channel, 10-Gbps Ethernet, and FCoE. Six fixed QSFP+ ports support 40-Gbps Ethernet and FCoE.

The chassis uses two 1+1 redundant, hot-swappable power supplies and three hot-swappable fans with 2+1 redundancy. Both airflow directions are supported.

Up to 128 buffer-to-buffer credits per port support Fibre Channel ISL operation. The datasheet specifies SAN extension up to 16 kilometers at 16-Gbps Fibre Channel speed with full Fibre Channel frames of 2122 bytes. Six 40-Gbps links can provide 240 Gbps of ISL bandwidth in an FCoE deployment.

Cisco Nexus 56128P

The 56128P is a 2RU chassis with 48 fixed 1/10-Gbps SFP+ ports and four fixed 40-Gbps QSFP+ ports. The base ports support 10-Gbps Ethernet and FCoE. The QSFP+ ports support 40-Gbps Ethernet and FCoE.

Two expansion slots accept the N56-M24UP2Q module. Each module supplies 24 unified SFP+ ports for 10-Gbps Ethernet, FCoE, or 2-, 4-, and 8-Gbps Fibre Channel, plus two 40-Gbps QSFP+ ports. The module is supported only in the 56128P.

The chassis includes four 2+2 redundant, hot-swappable power supplies and four 3+1 redundant, hot-swappable fans. Both port-side intake and fan-side intake configurations are supported.

Forwarding, buffering, and scale

The switches provide hardware Layer 2 and Layer 3 forwarding at line rate on all ports. The forwarding specifications are:

• Nexus 5672UP: 1.44 Tbps and 1071 mpps with 64-byte packets
• Nexus 5672UP-16G: 1.44 Tbps and 1071 mpps with 64-byte packets
• Nexus 56128P: 2.56 Tbps and 1904 mpps with 64-byte packets
• Up to 256,000 combined MAC address and ARP entries
• Approximately 1 microsecond port-to-port latency using cut-through forwarding
• 25 MB shared packet buffer per 12 x 10-Gbps SFP+ interfaces
• Up to 31 line-rate SPAN and ERSPAN sessions
• Up to 32,000 multicast routes in the platform feature description
• Up to 24 Layer 2 fabric extenders per switch

Presales sizing must account for the distinction between port count and effective uplink capacity. A 40-Gbps QSFP+ port configured as a breakout interface provides four 10-Gbps interfaces, but it does not provide four independent 40-Gbps paths. Oversubscription calculations should therefore be based on the intended server-to-uplink traffic pattern, east-west traffic, storage traffic, and failure scenarios.

The 25-MB shared buffer is allocated across groups of ports rather than as a dedicated buffer for every interface. Congestion-sensitive designs should map traffic classes, Fibre Channel or FCoE requirements, burst behavior, and uplink utilization to the relevant shared-buffer group.

Layer 2, Layer 3, QoS, and security

Layer 2 capabilities include IEEE 802.1Q VLANs, up to 4000 VLANs, VLAN trunks, PVRST+, MSTP with 64 instances, PortFast, root guard, Bridge Assurance, EtherChannel with up to 16 ports, LACP, vPC, vPC+, FabricPath, private VLANs, VLAN remapping, storm control, pause frames, Adapter FEX, and Data Center VM-FEX.

Jumbo frames are supported up to 9216 bytes. RoCE support uses Data Center Bridging, DCBX, and Priority Flow Control.

Layer 3 functions include routed ports, SVIs, port channels, subinterfaces, ECMP up to 64 ways, VRF-lite, IPv4 and IPv6 forwarding, BFD, HSRP, VRRP, policy-based routing, uRPF, and multicast. The stated scale includes:

• Up to 32,000 IPv4 host prefixes
• Up to 8000 IPv6 host prefixes
• Up to 8000 IPv4 multicast routes
• Up to 8000 IGMP snooping groups
• Up to 4000 VRF entries
• Up to 4096 VLANs
• Up to 4000 flexible ACL entries

Routing protocols include static routing, RIPv2, EIGRP, OSPFv2, BGP, IS-IS, OSPFv3, BGPv6, and EIGRPv6. Advanced routing functions such as BGP, full EIGRP, all PIM modes, IS-IS, uRPF, MSDP, sampled NetFlow, and VXLAN flood-and-learn require the applicable Layer 3 licensing.

QoS includes eight unicast queues and eight multicast queues per port, per-port QoS, CoS trust, ACL-based classification, MQC, virtual output queuing, strict-priority queuing, DWRR scheduling, flexible TCAM carving, and IPv4 and IPv6 CoPP.

Security functions include standard and extended Layer 2, Layer 3, and Layer 4 ACLs, PACLs, VACLs, RACLs, ingress policing, DHCP snooping with Option 82, Dynamic ARP Inspection, IP Source Guard, Ethernet port security, IPv6 ACLs, ACL logging for IPv4, and DHCP relay with up to 32 destinations.

FCoE and converged fabric

FCoE features require the Storage Protocols Services Package. Any supported 10-Gbps or 40-Gbps Ethernet port can be configured for FCoE where the hardware and software combination permits it.

The storage feature set includes Fibre Channel Forwarder operation, FIP, FCoE NPV, NPIV, VSAN trunking, FSPF, zoning, enhanced zoning, fabric binding, Fibre Channel-SP authentication, FCID persistence, device alias services, RSCN, name-server services, in-order delivery, port tracking, direct FCoE target attachment, boot from SAN over vPC and EvPC, Fibre Channel traceroute, Fibre Channel ping, and Fibre Channel debugging.

The 5672UP-16G is the preferred model where 16-Gbps Fibre Channel is required on the switch itself. The 5672UP supports Fibre Channel on its last 16 unified ports at 2-, 4-, and 8-Gbps speeds. The 56128P requires the unified expansion module for the Fibre Channel interface set described in the datasheet.

Fabric Extender and fabric designs

The Nexus 5600 switches can operate as parent switches for Cisco Nexus 2200 and 2300 Fabric Extenders. Supported deployment modes include single-connected and dual-connected FEX designs using vPC+. FEX provides a single management plane and supports rack-server, blade-server, virtualized access, mixed 1/10-Gbps access, 10GBASE-T migration, FCoE, and high-density aggregation scenarios.

Supported fabric extenders include:

Leaf-and-spine designs can use the Nexus 5600 switches as either leaf or spine nodes. Presales designs should validate the required number of leaf uplinks, ECMP width, vPC requirements, FEX attachment model, multicast scale, and storage traffic isolation before selecting the chassis.

Power, environmental, and physical requirements

The platform supports front-to-back and back-to-front airflow. Airflow direction must match the rack cooling design. Mixing airflow directions within a rack row can reduce cooling effectiveness and should be avoided.

Power supplies are hot-swappable and support redundant configurations. The available supply families are:

Typical operating power is 650 W for each supply family. Heat dissipation is listed as 45 BTU/hr for the N55-PAC-1100W and NXA-PAC-1100W, 260 BTU/hr for the N55-PDC-1100W, and 170 to 130 BTU/hr for the NXA-PHV-1100W.

No PoE feature or PoE power budget is specified. These switches should not be proposed as Power over Ethernet access switches.

Transceiver and cabling guidance

Supported 10-Gbps options include FET-10G, SFP-10G-SR, SFP-10G-SR-S, SFP-10G-LR, SFP-10G-LR-S, SFP-10G-ER, SFP-10G-ER-S, SFP+ Twinax cables from 1 to 5 meters, active Twinax cables at 7 and 10 meters, and active optical cables from 1 to 10 meters.

Supported 40-Gbps options include SR4, CSR4, LR4, LR4-S, LR4 Lite, BiDi, and ER4 optics, plus direct-attach copper, active copper, direct-attach optical, and breakout assemblies. Optical distances include 100 meters for SR4 over OM3, 150 meters for SR4-S over OM4, 300 meters over OM3 and 400 meters over OM4 for CSR4, 10 kilometers for LR4, 2 kilometers for LR4 Lite, and 40 kilometers for ER4.

QSFP+ breakout cables and the CVR-QSFP-SFP10G1 adapter allow 40-Gbps ports to connect to 10-Gbps Ethernet optics. The QSFP-to-four-SFP+ option should be included when a design requires four 10-Gbps links from one QSFP+ port.

Software and licensing

NX-OS provides CLI, SNMP, XML and NETCONF management, configuration rollback, configuration synchronization, RBAC, AAA, RADIUS, TACACS+, SSHv2, SPAN, ERSPAN, embedded packet analysis, EEM, GOLD diagnostics, Call Home, and DCNM integration.

The Layer 3 Base Services license is required before using the Layer 3 Enterprise Services license. FCoE features require the storage services license.

Ordering matrix

Cisco specifies a one-year limited hardware warranty with hardware replacement on a 10-day turnaround after receipt of an RMA. Available service offerings include Advanced Services, SMARTnet, Smart Call Home, and operational support for deployment, optimization, troubleshooting, migration, and lifecycle management.