Elastic Networks BitStorm L3S-T, Руководство пользователя, Страница: 9 / 167

Stacking BitStorm L3S Switches
As new networks are deployed and existing ones continue to grow aggressively, managers need cost-effective
products that can adapt. Low cost, high-performance “stacked” switches that can be managed as a single entity
are the most desirable solutions, especially for small to medium-sized enterprises.
What is a “stackable” switch?
A stackable solution ensures that a “master” switch can be connected to one or more “slave” switches and that all
can function or be managed as a single logical device.
Built in a predominately standalone fixed-port configuration, this type of switch is typically a single-board system
that is self-contained in an enclosure with its own power supply.
Port density is increased by connecting one switch to another, unlike a chassis-based system in which ports are
added using expansion boards. A stackable switch is connected in a peer-to-peer or in a master-slave relationship
to switches of equal or similar size.
When a “stack” is not a stack
Many manufacturers today say their switches are “stackable” simply because they can be connected using a
single Gigabit uplink on each switch.
This not only “burns” switch resources by stealing a Gigabit link, it is an ineffective design that creates severe
blocking and packet loss between switches. Others use a “virtual chassis” where a separate switch is used as a
“traffic cop” to interconnect switches, again, using gigabit ports for this connection and creating both non
wire-speed transfers and blocking.
These switches are more accurately described as “linked” not “stacked”. They cannot truly be called stackable
switches because:
these external Gigabit links introduce a significant degree of blocking
●
these connections consume switch ports
●
in many cases, they are not necessarily managed as a single unit and therefore cannot truly be called
stackable switches.
●
Linking
Some manufacturers “stack” their switches
by using a Gigabit uplink on each switch.
This uses valuable resources and creates
severe blocking and packet loss.
Virtual Chassis
Other manufacturers use a “virtual chassis” concept that burns
Gigabit ports while creating non wire-speed transfers and
blocking constraints.
Stacking
The BitStorm L3S system uses dedicated 8
Gbps stacking interfaces to guarantee
wire-speed, non-blocking performance.
When a stack is a stack
The only true stacking interface is a design that uses an external, high-speed bus to interconnect separate
stackable switches.
High-speed stacking bus
This interface connects the ports on the stackable “slave” switches directly to the switching fabric on the “master”
switch. This is the only switch interface technology that can deliver the bandwidth necessary for wire-speed packet
forwarding and eliminate blocking between connected switches.
What is blocking?
Basically, blocking is the inability of a switch to forward traffic due to bandwidth limitations. Technically, packet
transfers are blocked when sufficient bandwidth is not available for all packets to be forwarded at the highest
speed possible on the link.
Packets can be blocked externally as they are forwarded between switches as well as internally within the switch.
Internal, or head-of-line blocking, is eliminated through complex buffering and queuing, while blocking between
9