Beruflich Dokumente
Kultur Dokumente
Victor Lama
Fabric Specialist LE Northeast Region
August 2010
What if?
What if your company could drastically minimize the complexity of managing your data center network by
being able to take multiple network appliances and manage them as if they were just one device?
What if there was a modular network switching solution that offered dual routing and switching engines with
graceful failover in the event of an outage, as well as in-service hardware replacement and code upgrades?
What a tremendous impact that would make toward providing high availability to mission critical applications
and services!
What if network architects could finally do away with the design complexities of deploying the Spanning Tree
Protocol by virtualizing multiple layers in the data center? No more worries about interoperability of STP,
MSTP, RSTP, PVRSTP, CST, across vendor platforms, and no more blocking and wasting half your data centers
cross-sectional bandwidth!
What if those who hold your companys purse strings could scale your network CAPex as the business need
arises, instead of just having to make large capital investments up front?
With Junipers Virtual Chassis technology, all the above, and more, are possible.
intelligence of all the members of the VC, thereby ensuring that traffic takes the shortest path to its
destination.
VC is one of the core enabling technologies of Junipers vision for the data center of the future, otherwise
known as Stratus. The foundational concept behind Stratus is to replicate the simplicity of a single fabric in a
single switch chassis across the data center or between data centers! The result is a resilient low-latency
data center fabric that has the built in flexibility to support converged networks and virtualized workloads in a
variety of different topologies.
Once the VCP ports are cabled, they are automatically enabled to provide the switching backplane for the
dedicated chassis configuration.
To extend the VC configuration across a campus or regional area, the EX-UM-2X4SFP 10Gbps module, with
dual 10G ports, can be deployed and configured to carry VC control information over an aggregated link that
provides up to 40Gbps of throughput 20Gbps in each direction.
Cabling Options
There are three different methods for connecting members of a Virtual Chassis. These are the daisy-chained
ring, the braided-ring and the extended configuration. The daisy-chained and braided ring methods are
deployed in dedicated chassis configurations, with the braided-ring offering a greater maximum distance
between the first and last member of the stack.
As noted earlier, an extended chassis consists of smaller dedicated virtual chassis interconnected by Ethernet
ports in a singular or bundled fashion. This capability provides the flexibility to manage an entire campus LAN
as one management plane with one IP address and a singular control plane that allows for automatic
configuration and code upgrades upon introduction of a new member to the VC. This is known as dynamic
installation.
See figure 6C below.
As figure 8 shows, in a Dell m1000e blade server deployment that includes M-series stackable blade switches,
virtualization of the edge and access layer can be achieved, thereby removing the need for the deployment of
the Spanning Tree Protocol. Multiple physical connections between the blade switches and the ToR VC will be
seen as one logical link by each virtual layer. This allows for a non-blocking rack architecture and full utilization
of available cross-sectional bandwidth.
Summary
Deploying a Virtual Chassis solution in a high performance data center reduces the complexity of switch
hardware installations, OS upgrades, configuration tasks, and management of the switched fabric. Moreover,
virtualizing multiple layers is one of the necessary steps toward the creation of a flat data center network in
which all L2 links are utilized and multiple layers of packet processing are eliminated. This will minimize
latency, remove the need to deploy L2 loop mitigation technology, and create a non-blocking fabric with an
average x2 increase in available cross-sectional bandwidth across the data center.