User's Manual

Design Guide

●

Network Fault Tolerance (NFT)

●

Transmit Load Balancing (TLB)

●

Switch Assisted Load Balancing (server load balancing [SLB])

NFT teaming creates a virtual interface by grouping the blade-server network adapters into a team.

One adapter is the primary active interface and all other adapters are in a standby state. The virtual

adapter uses a single MAC address and a single Layer 3 address. NFT provides adapter fault

tolerance by monitoring the state of each team member’s network connection. The standby NICs

become active only if the primary NIC loses connectivity to the network.

TLB teaming supports adapter fault tolerance (NFT) and adds more functions in the server for load

balancing egress (transmit) traffic across the team. Note that a TLB team uses only one NIC to

receive traffic. The load-balancing algorithm is based on either the destination MAC or IP address.

This teaming method provides better use of the bandwidth available for egress traffic in the network

than NFT.

SLB teaming extends the functions of TLB by allowing the team to receive load-balanced traffic

from the network. This reception requires that the switch can load balance the traffic across the

ports connected to the server NIC team. The Cisco Catalyst Blade Switch 3020 supports the IEEE

802.3ad standard and Gigabit Ethernet port channels. SLB teaming can only be used on full-height

servers, because it requires that both NICS go to the same upstream switch.

For more information about NIC teaming, please visit:

http://h18000.www1.hp.com/products/servers/networking/whitepapers.html

Scalability

The capability of the data center to adapt to increased demands without compromising its

availability is a crucial design consideration. The aggregation layer infrastructure and the services it

provides must accommodate future growth in the number of servers or subnets it supports.

When deploying blade servers in the data center, two primary factors need to be considered:

●

Number of physical ports in the aggregation and access layers

●

Number of slots in the aggregation layer switches

Physical Port Count

The introduction of blade systems into the data center requires greater port density at the

aggregation layer. Blade systems, deployed with internal switches, provide their own access layer.

The cabling and maximum number of servers per enclosure are predetermined. Scaling the

aggregation layer ports to accommodate the blade-system uplinks is an area that requires

attention.

It is important to remember that aggregation switches provide data center services such as load

balancing, security, and network analysis that may require dedicated ports for appliances or slots

for integrated services. This situation directly affects the number of ports available for access layer

connectivity.