Local switching allows neighboring ports to communicate without consuming valuable core / slot bandwidth. It also results in lower switching latency and higher full-speed port density.
How does Brocade compare?
Brocade is the only vendor that delivers local switching. In the 48000, in addition to 64 Gbit/sec of core switching bandwidth per slot,
- the 4 Gbit/sec 32-port director blade provides 128 Gbit/sec of local switching -- if all 64 Gbit/sec of slot bandwidth is used, 64 Gbit/sec of local switching is still available on the blade.
- the 4 Gbit/sec 48-port director blade has 192 Gbit/sec of local switching -- if all 64 Gbit/sec of slot bandwidth is used, 128 Gbit/sec of local switching is still available on the blade.
- 8 Gbit/sec blades will support local switching in the 48000
In the DCX, in addition to 256 Gbit/sec of core switching bandwidth per slot,
- the 8 Gbit/sec 48-port director blade has 384 Gbit/sec of local switching -- if all 256 Gbit/sec of slot bandwidth is used, 128 Gbit/sec of local switching is still available on the blade.
- there is no oversubscription in slots when using either 16 or 32 8 Gbit/sec blades, but local switching bandwidth is still available.
- there is no oversubscription in slots when using 4 Gbit/sec optics in the 48 port 8 Gbit/sec blades, and local switching bandwidth is still available.
As a result, Brocade has the industry's highest full-speed 4 Gbit/sec and 8 Gbit/sec port density.
| Function | Brocade DCX | Brocade 48000 | Brocade Mi10000 | Brocade M6140 | Cisco MDS 9513
| Local Switching
| Yes | 16 to 48 x 8 Gbit/sec ports Yes | 16 to 48 x 4 or 8 Gbit/sec ports No | No | No
| Local bandwidth per blade
| 128 to 384 Gbit/sec
| 64 to 192 Gbit/sec
| 0 | 0 | 0
| Backplane bandwidth per slot
| 256 Gbit/sec | (32 x 8 Gbit/sec) (32 x 4Gb + 16 x 8Gb) (&#lt; 48 x 4 Gbit/sec) 64 Gbit/sec | (16 x 4 Gbit/sec) 64 Gbit / sec | (16 x 4 Gbit/sec) 8 Gbit/sec | (2 x 4 Gbit/sec) 48 Gbit/sec | (12 x 4 Gbit/sec) |
|---|
Cisco MDS directors cannot switch locally. Traffic between neighboring ports must use valuable backplane bandwidth that is already at a premium with only 51.2 Gbit/sec of bandwidth per linecard slot. As a result, no more than 12 ports per slot can simultaneously communicate at 4 Gbit/sec, even when the 24 and 48-port linecards are in place. While ports on the 24 and 48-port linecards no longer need to be turned off when the 12 ports are set at 4 Gbit/sec, they must share a meagre 3.2 Gbit/sec, or 0.8 Gbit/sec per port group.*.
