Full English Description for Cisco N9K-C93600CD-GX
1. Part Number Naming Breakdown
-
N9K: Cisco Nexus 9000 cloud-scale fixed data center switch family
-
C93600CD: Hybrid dual-port leaf chassis mixing QSFP28 downlinks + QSFP-DD 400G uplinks, 1RU compact form factor
-
C: QSFP28 multi-rate optical ports for server/GPU/storage downlink access
-
D: QSFP-DD dual-density high-speed optical ports supporting native 400G fabric uplinks
-
GX: First-generation GX Cloud Scale ASIC platform, large shared packet buffer, expanded L2/L3 TCAM scale, hardware RoCEv2 lossless transport, dual independent operating modes: standalone NX-OS and ACI intent-based fabric mode
2. Short Sales One-Liner (For BOM, Quotation & Datasheet Header)
Cisco Nexus N9K-C93600CD-GX, 1RU rack-mount hybrid high-speed leaf data center switch, 28 × 10/40/100G QSFP28 downlink ports plus 8 × 100/400G QSFP-DD fabric uplink ports, 6.0 Tbps non-blocking switching capacity, dual hot-swappable 1100W Platinum-efficiency redundant power supplies and six 5+1 N+1 variable-speed fan trays, dual-mode standalone NX-OS or native ACI fabric operation, purpose-built for AI GPU training ToR leaf aggregation, HPC low-latency clusters, converged compute/storage fabrics, multi-pod ACI deployments and hybrid 100G/400G hyperscale cloud leaf-spine fabricsCisco.
3. Complete Product Overview Paragraph
The Cisco Nexus N9K-C93600CD-GX is a space-saving 1RU fixed-port Layer 2/3 hybrid optical leaf switch built on Cisco’s first-generation GX Cloud Scale ASIC, purpose-engineered as top-of-rack leaf nodes for modern AI/ML GPU clusters, high-performance computing (HPC), virtualized enterprise storage and Cisco ACI Application Centric Infrastructure multi-pod environmentsCisco. It delivers a balanced mixed-speed port layout optimized for converged workloads: 28 universal QSFP28 ports dedicated to GPU, server and storage downlinks, paired with eight high-bandwidth QSFP-DD dual-density ports for high-speed 400G spine interconnection, delivering full wire-rate non-blocking forwarding with total switching bandwidth of 6.0 Tbps and forwarding throughput of 4.5 Bpps for minimum 64-byte Ethernet frames.
All 28 QSFP28 downlink ports support native 40G and 100G transceivers with flexible breakout capabilities, subject to fixed four-port MAC quad grouping rules: ports 1–24 are divided into sequential 4-port quads, and all ports within one quad must run identical speed/breakout configurations (mixed speeds inside a quad are hardware blocked). Valid breakout options for QSFP28 ports include 4×10G, 4×25G, 4×50G, 2×100G via breakout DAC cables and QSA adapters for backward 10G connectivity. The eight QSFP-DD uplink ports (ports 29–36) natively support 400GBASE, backward-compatible 100G QSFP28 transceivers, with rich breakout profiles: 2×200G, 4×100G, 4×50G, 4×25G, 4×10G, ideal for seamless fabric migration from 100G to 400G core inter-switch links. All QSFP-DD uplink ports support wire-rate 128/256-bit MACsec encryption to secure inter-pod fabric connectionsCisco.
Powered by GX-series advanced Cloud Scale silicon, this platform features a large shared dynamic packet buffer to absorb heavy burst traffic generated by parallel GPU training workloads, drastically expanded MAC address table and ultra-large IPv4/IPv6 LPM TCAM route scale for hyperscale multi-tenant segmentation, native hardware acceleration for EVPN/VXLAN overlay routing, and deeply optimized RoCEv2 lossless congestion control with PFC for ultra-low-latency HPC and AI clusters. High-performance unsampled streaming telemetry integrates fully with Cisco Tetration Analytics for end-to-end full-fabric observability, alongside expanded VRF, multicast and microsegmentation policy capacity for isolated multi-tenant cloud workloads. The chassis ships with factory default leaf port partitioning and pre-tuned POAP zero-touch provisioning workflows to simplify ACI fabric deployment and APIC controller discovery.
The chassis integrates dual independent out-of-band management interfaces: dedicated 10/100/1000BASE-T RJ45 Ethernet and SFP multi-rate fiber remote management uplink ports, plus mini-USB RS232 local console and single USB 2.0 Type-A port to support offline configuration backup and firmware upgrades. Dual hot-swappable 1100W power supplies and six N+1 redundant fan trays eliminate single points of hardware failure, with two standardized airflow variants (PI port-side intake / PE port-side exhaust) selectable at order time; all power and fan modules must match airflow direction during operation to avoid thermal overload. This switch supports two independent operating modes: modular standalone Cisco NX-OS for traditional routed data center architectures, or native ACI fabric mode for policy-driven automated multi-tenant cloud infrastructure. Its compact 1RU hybrid 100G/400G design maximizes rack density for converged GPU compute and storage leaf deployments with high-speed 400G uplink scalability to the fabric spine layerCisco.
4. Detailed Hardware Interface Specifications
QSFP28 Downlink Ports (28 × Multi-Rate Slots, Ports 1–28)
-
Native supported line rates: 40GBASE, 100GBASE
-
Quad grouping constraint: Ports 1–24 split into fixed 4-port MAC quads; uniform speed/breakout required per quad
-
Supported breakout profiles: 2×100G, 4×50G, 4×25G, 4×10G via breakout cables + QSA adapters
-
Transceiver compatibility: QSFP28 100G SR4/LR4/ER4, QSFP+ 40G SR4/LR4, SFP28 25G, SFP+ 10G via QSA, passive QSFP28 DAC twinax cables
QSFP-DD Fabric Uplink Ports (8 × Multi-Rate Slots, Ports 29–36, all MACsec-capable)
-
Native supported line rates: 100GBASE, 400GBASE
-
Full independent breakout options: 2×200G, 4×100G, 4×50G, 4×25G, 4×10G
-
Transceiver compatibility: QSFP-DD 400G DR4/FR4/LR4/ZR, QSFP28 100G, passive QSFP-DD DAC twinax cables
-
Primary use case: 400G high-bandwidth uplinks connecting leaf to spine switches in leaf-spine fabrics
Management & Auxiliary Ports
-
1 × 10/100/1000BASE-T RJ45 out-of-band management port
-
1 × SFP multi-rate fiber remote out-of-band management uplink port
-
1 × Mini-USB RS232 console port for local CLI access
-
1 × USB 2.0 Type-A port for firmware upload, configuration backup and POAP deployment
Power & Cooling
-
Dual hot-swappable power supply options (1+1 load-sharing redundancy, non-disruptive hot-swap):
1100W AC PI, 1100W AC PE, 1100W DC PI, 1100W HVAC/HVDC PI/PE
-
Six variable-speed fan trays (5+1 N+1 redundancy):
NXA-SFAN-35CFM-PI (burgundy label, port-side intake)
NXA-SFAN-35CFM-PE (blue label, port-side exhaust)
-
Power efficiency: 80Plus Platinum certified
-
Operating temperature range: 0°C ~ 40°C full load; non-operating storage: -40°C ~ 70°C
-
Humidity tolerance: 5%–95% non-condensing; max operating altitude: 4000 meters
Physical Form Factor
-
Dimensions: 1RU standard 19-inch rack chassis
Height: 4.37 cm (1.72 in), Width: 44.13 cm (17.37 in), Depth: 64.8 cm (25.5 in)
-
Empty weight (no PSUs/fans): ~8.9 kg; fully populated weight ~11.6 kg
-
Compliance: RoHS-6, NEBS Level 3, global safety, EMC and telecom regulatory certifications
-
MTBF: 295,515 hours under standard operating conditions
Core Hardware Performance Benchmarks
-
Total switching capacity: 6.0 Tbps full duplex non-blocking
-
Forwarding rate: 4.5 Bpps (64-byte frame wire-rate)
-
Shared dynamic packet buffer: Large GX-series shared buffer pool for cross-port burst traffic absorption
-
MAC address table size: Expanded GX-scale large MAC forwarding entry pool
-
IPv4 / IPv6 unicast LPM route table capacity: Ultra-large TCAM scale for hyperscale multi-pod fabrics
-
VLAN support: Up to 4094 standard IEEE 802.1Q tagged VLANs
-
Maximum isolated VRF instances: 16,000 multi-tenant virtual routing tables
-
System memory: 32GB DRAM + 128GB high-capacity onboard SSD storage
-
Typical cut-through switching latency: Less than 1 microsecond for minimum-size Ethernet frames
5. Core Software Feature Suite (Dual NX-OS / ACI Operating Modes)
Layer 2 & Layer 3 Routing
Static routing, RIP, OSPFv2/v3, EIGRP, full BGP routing suite, VRRP, PIM sparse/dense multicast; limited FCoE support via designated QSFP28 downlink ports for unified storage fabrics
Cloud Fabric Overlay
Hardware-accelerated EVPN/VXLAN symmetric & asymmetric routing, multi-tenant VXLAN segmentation, fabric multicast optimization, advanced lossless RoCEv2 congestion control optimized for GPU/AI/HPC low-latency training workload clusters
Enterprise & Fabric Security
802.1X port authentication, MAB MAC bypass authentication, DHCP Snooping, Dynamic ARP Inspection, CoPP control-plane policing, hardware wire-rate 128/256-bit MACsec encryption on all 8 QSFP-DD fabric uplink ports, TrustSec SGT zero-trust micro-segmentation for core fabric isolation
Monitoring & Telemetry
High-performance sampled/unsampled IPFIX streaming telemetry, ERSPAN remote port mirroring, local SPAN/VSPAN, real-time per-port/per-queue buffer utilization visibility, full Cisco Tetration flow export compatibility for end-to-end fabric analytics
ACI Default Profile Built-In Features
Factory predefined leaf port partition (28 downlinks + 8 fabric uplinks), native APIC controller fast discovery, streamlined ACI POAP zero-touch provisioning workflow, built-in inter-pod IPN link validation templates
Automation & DevOps
POAP zero-touch provisioning, NX-API REST JSON/XML APIs, embedded Linux Bash shell, native Python scripting, Ansible/Puppet/Chef full integration, containerized application hosting, hot patching and ISSU in-service software upgrade without leaf fabric traffic downtime
High Availability
Modular fault-isolated process architecture, stateful process restart, redundant control plane, hot-swappable power supply and fan FRUs for non-disruptive field maintenance
6. Typical Deployment Use Cases
-
Top-of-Rack leaf switch for converged 100G GPU virtualization, AI model training and high-speed storage racks with 400G spine uplinks
-
Low-latency leaf aggregation node for HPC parallel computing and big-data analytics workload clusters
-
Converged compute/storage fabric leaf supporting FCoE block storage traffic for enterprise unified data center architectures
-
Dedicated leaf access node for medium and large-scale ACI multi-pod Application Centric Infrastructure fabrics
-
Hybrid migration leaf for mixed-generation 100G server environments with upgradeable 400G high-bandwidth uplink capacity to spine layer
-
Hyperscale IaaS multi-tenant cloud leaf optimized for balanced compute/storage workload density and future 400G fabric expansion
7. Datasheet Short Summary Paragraph
The Cisco Nexus N9K-C93600CD-GX is a compact 1RU GX-generation hybrid optical cloud data center leaf switch equipped with 28 breakout-capable universal QSFP28 40/100G downlink ports and eight multi-rate 100/400G QSFP-DD fabric uplink ports, delivering 6.0 Tbps non-blocking switching performance with dual 1100W Platinum redundant power and six 5+1 redundant fan trays. Supporting both standalone modular NX-OS and native ACI intent-based fabric operational modes, it features the first-gen GX Cloud Scale ASIC with large shared packet buffer, ultra-expanded L2/L3 TCAM routing scale, wire-rate MACsec encryption on all 400G uplink ports and optimized RoCEv2 lossless transport, serving as a balanced hybrid-speed leaf aggregation switch for AI/GPU HPC, converged storage, multi-pod ACI fabrics and hybrid 100G/400G hyperscale cloud data centers.
|
