22.05.2017 Views

nx.os.and.cisco.nexus.switching.2nd.edition.1587143046

Nexus Switching 2nd Edition

Nexus Switching 2nd Edition

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

----------------<br />

Ethernet1/1 TBM12224047 1 Ethernet1/2 b<br />

Ethernet1/3 TBM12224047 1 Ethernet1/4 b<br />

Cisco FabricPath<br />

Historically, Layer 2 networks did not require high levels of scalability because network<br />

administrators designed networks based on the limitations of spanning tree, hardware MAC<br />

address tables, <strong>and</strong> L2 flooding characteristics. Doing so limited Layer 2 to small pockets<br />

<strong>and</strong> separated them from one another via L3 routing devices. Leveraging L3 boundaries<br />

around these small pockets of L2 infrastructure allowed network administrators to rely on<br />

more scalable resilient routing protocols to scale their networks. Trends today in the areas<br />

of high-performance computing (HPC), virtualization, <strong>and</strong> a dem<strong>and</strong> for a more flexible<br />

infrastructure requires many organizations to revisit Layer 2 technologies. New technologies<br />

such as virtual Port Channel (vPC) are emerging to help customers build larger L2 domains;<br />

however, these technologies still fundamentally rely on spanning tree. Spanning tree creates<br />

topologies where local problems such as a link flap can have catastrophic impacts to the<br />

entire infrastructure, whereas more intelligent Layer 3 routing protocols such as OSPF <strong>and</strong><br />

EIGRP assume these types of failures <strong>and</strong> have sophisticated algorithms to h<strong>and</strong>le such<br />

events. These challenges have created the dem<strong>and</strong> for a more flexible protocol, which takes<br />

the best characteristics of Layer 2 <strong>and</strong> Layer 3 <strong>and</strong> combines them to meet these dem<strong>and</strong>s.<br />

Cisco FabricPath introduces a new way to build <strong>and</strong> operate Layer 2 networks. FabricPath<br />

enables network operators to build large-scale L2 networks <strong>and</strong> eliminate challenges<br />

associated with various spanning tree implementations. FabricPath takes well-understood<br />

technologies that have been leveraged at Layer 3 for routing IP packets <strong>and</strong> uses them to<br />

carry Layer 2 reachability information. In this manner, network administrators gain the<br />

flexibility of large L2 networks <strong>and</strong> still have the resiliency of large L3 networks.<br />

Cisco FabricPath can be broken down into two distinct areas of operations: control plane<br />

<strong>and</strong> data plane. The control plane is responsible for building <strong>and</strong> disseminating information<br />

about the reachability of devices within the network, whereas the data plane is responsible<br />

for the encapsulation, de-encapsulation, <strong>and</strong> forwarding of data packets acr<strong>os</strong>s a FabricPath<br />

network.<br />

The FabricPath control plane uses Intermediate System-to-Intermediate System (IS-IS) to<br />

replace spanning tree. IS-IS is a multi-area (level) routing protocol that has no IP<br />

dependencies <strong>and</strong> is easily extensible using custom TLVs. These characteristics make IS-IS<br />

an attractive way to exchange routing information at Layer 2. The FabricPath control-plane<br />

functions can be broken down into two main areas. First, the FabricPath domain is<br />

comp<strong>os</strong>ed of a number of switches, which must be uniquely identified. The FabricPath<br />

domain may also consist of a number of disparate logical topologies for isolation of various<br />

types of traffic. Dynamic Resource Allocation Protocol (DRAP) provides the mechanism<br />

for switches to be auto-assigned a unique 12-bit switch ID (SID). This protocol is built on<br />

IS-IS adjacencies, which are formed when an interface first connects to another FabricPath<br />

switch. In addition, the network administrator may cho<strong>os</strong>e to statically configure the SID.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!