Elastic Networks BitStorm L3S-T, Руководство пользователя, Страница: 10 / 167

switches is a simple matter of providing sufficient bandwidth for wire-speed packet transfers.
How BitStorm L3S stacking eliminates blocking
BitStorm L3S
's
high performance stacking architecture guarantees that packets are forwarded at wire speed to all
ports on all switches in the stack without blocking any transmissions. BitStorm L3S does this using a dedicated
high-speed interface connecting all ports directly to the central switching fabric.
Blocking between switches
To avoid blocking between switches, a stackable switch must be able to forward the full traffic load from any of its
switch ports to any switch port on any switch in its stack.
Using the example of a single Gigabit uplink that is used to connect two switches with 24 Fast Ethernet ports, that
single Gigabit uplink is
less than half the bandwidth needed to prevent blocking
between two switches.
At full duplex, that single uplink delivers only 2Gbps of bandwidth instead the 4.8Gbps needed to forward packets
at wire-speed over all 24 Fast Ethernet ports also operating at full duplex.
At full duplex, 24 Fast Ethernet ports talking to 24 Fast Ethernet ports need 200 Mbps x 24,
or 4.8 Gbps of bandwidth for non-blocking performance.
At full duplex, a single Gigabit link between switches only provides 2 Gbps of bandwidth,
less than half of what's required.
BitStorm L3S stacking dedicates 8Gbps of bandwidth between 24-port switches,
almost double the 4.8 Gbps required for non-blocking transfers.
Single entity management
In a BitStorm L3S stack, the management software running in the master extends its power over the ports on the
slave switches. In effect, the slaves rely upon the greater power of the master. As such, slaves are very cost
effective. The master CPU runs a single management system that sees all ports in the stack as its own, making
the slave switches transparent to the network.
10