Cisco Cisco Nexus 5010 Switch Guida Alla Progettazione

 

Design Guide 

 

© 2010 Cisco Systems, Inc. All rights reserved. This document is Cisco Public Information. 

Page 6 of 22

 

● 

Architectures that provide an intrinsic lower latency than traditional LAN networks, so that computing cloud 
can be built on the same LAN infrastructure as regular transactional applications. 

● 

Architectures that provide the ability to distribute Layer 2 traffic on all available links. 

● 

Simplified cabling: For a more efficient airflow, lower power consumption, and lower cost of deployment of 
high-bandwidth networks. 

● 

Reduction of management points: It’s important to limit the impact of the sprawl of switching points (software 
switches in the servers, multiple blade switches, and so on). 

The next-generation data center provides the ability to use all links in the LAN topology by taking advantage of 
technologies such as virtual PortChannels (vPCs). vPCs enable full, cross-sectional bandwidth utilization among LAN 
switches, as well as between servers and LAN switches. 

Most rackable servers today include redundant LAN-on-motherboard (LOM) interfaces for management, an 
integrated-lights-out (iLO) standard-based port, and one or more Gigabit Ethernet interfaces, and redundant host bus 
adapters (HBA). The adoption of 10 Gigabit Ethernet on the server simplifies server configuration by reducing the 
number of network adapters and providing enough bandwidth for virtualized servers. The data center design can be 
further optimized with the use of Fibre Channel over Ethernet (FCoE) to build a unified fabric. 

Cost-effective 10 Gigabit Ethernet connectivity can be achieved by using copper twinax cabling with Small Form-
Factor Pluggable Plus (SFP+) connectors. 

A rackable server configured for 10 Gigabit Ethernet connectivity may have an iLO port, a dual- LOM, and a dual-port 
10 Gigabit Ethernet adapter (for example, a converged network adapter). This adapter would replace multiple Quad 
Gigabit Ethernet adapters and, in case the adapter is also a Cisco Network Adapter, it would also replace an HBA. 

Fabric extender technology simplifies the management of the many LAN switches in the data center by aggregating 
them in groups of 10 to 12 under the same management entity. In its current implementation, Cisco Nexus 2000 
Series Fabric Extenders can be used to provide connectivity across 10 to 12 racks that are all managed from a single 
switching configuration point, thus bringing together the benefits of top-of-the-rack and end-of-the-row topologies. 

Significant cost reduction can be achieved by replacing Quad Gigabit Ethernet cards and dual HBAs with a dual-port 
converged network adapter card connected to a data-center-bridging (DCB)-capable device. A device such as the 
Cisco Nexus 5000 Series Switch also provides Fibre Channel Forwarder (FCF) capability.  

The next-generation data center provides no-drop capabilities through priority flow control (PFC) and a no-drop 
switch fabric for suitable traffic types. 

While not mandatory, using a cut-through-capable access layer enables low-latency communication between servers 
for any packet size. The Cisco Nexus 5000 Series supports deterministic 3.2- microseconds latency for any packet 
size with all features enabled (with access control list [ACL] filtering applied, as an example) and similarly the Cisco 
Nexus 4000 Series supports 1.5- microseconds latency.  

Figure 2 illustrates how a next-generation data center would look like with Cisco Nexus Switches and vPC.