110 Brocade DCX 8510-4 Backbone Hardware Reference Manual
53-1002177-05

System architecture

A
System architecture

Table 14 describes the system architecture of the Brocade DCX 8510-4.

Port to port latency Local switching - 800ns (add 400ns for E_Port to E_Port);
Blade to blade - 2.4 microseconds (add 400ns for E_Port to E_Port)
Switching capacity An aggregate switching capacity of 3.36 billion frames per second (for Class 2,
Class 3, and Class F frames for a 192 port chassis
TABLE 13 General specifications (Continued)
TABLE 14 System architecture
Feature Description
Fibre Channel Ports Up to 192 16 Gbps ports, universal (E_Port, F_Port, EX_Port, M_Port, D_Port).
Control Processor Redundant (active/standby) control processor blades.
Scalability Full fabric architecture: 239 switches maximum.
Performance 2.125 Gbps line speed, full duplex.
4.25 Gbps line speed, full duplex.
8.50 Gbps line speed, full duplex.
10.51875 Gbps line speed, full duplex.
16.0 Gbps line speed, full duplex.
Autosensing of 2, 4, 8, and 16 Gbps port speeds depending on SFPs used. Speed
matching between 2, 4, 8, and 16 Gbps port speeds. 10 Gbps port speeds with
dedicated SFPs.
ISL Trunking Automatically uses up to 4 ports in a trunk group to form a 32 Gbps trunk.
Can use up to 8 ports in a trunk group to form a 64 Gbps trunk.
Aggregate bandwidth 4.1 Tbps populated with four FC16-48 port blades and using 16 ICL ports @ 64 Gbps).
Switch latency FC8-64 blade:
<2.1 µsec any port to any port at 8 Gbps, cut-through routing.
FC8-32E blade:
<800 ns any port to any port, 2.4 µsec blade to blade at 8 Gbps, cut-through routing.
FC8-48E blade:
<800 ns any port to any port, 2.4 µsec blade to blade at 8 Gbps, cut-through routing.
FC16-32 blade:
<800 ns any port to any port, 2.4 µsec blade to blade at 16 Gbps, cut-through routing.
FC16-48 blade:
<800 ns any port to any port, 2.4 µsec blade to blade at 16 Gbps, cut-through routing.
FX8-24 blade - FC to FC:
<2.1 µsec any port to any port at 8 Gbps, cut-through routing.
Maximum frame size 2112-byte payload.
Frame buffers 8000 per ASIC, with 5000 more that can be borrowed from neighboring ASIC if
enabled.