HP Integrity rx2800 HP Integrity Network Adapter Teaming Whitepaper - Page 70

As an example, a detailed discussion of Cisco’s EtherChannel load-balancing algorithm is

described below. Most of the information about how Cisco switches load-balance traffic on a

port trunk is applicable to other switch vendors’ Port Trunking technology.

Cisco EtherChannel

Cisco's Fast EtherChannel (FEC) and Gigabit EtherChannel (GEC) technology is a MAC layer

(Layer 2) load-balancing technology using two or more switch ports grouped together as one

logical switch port. Depending on the specific load-balancing algorithm used, FEC/GEC may

not efficiently load balance traffic to an SLB team.

FEC/GEC was originally designed as a switch-to-switch technology allowing two switches to

increase the bandwidth between each other by aggregating multiple ports together as a single

logical port for both transmits and receives. This is in contrast to TLB that only balances transmit

traffic. An algorithm had to be used that could statistically divide the traffic over each port in

the FEC/GEC group in an attempt to divide it evenly.

There are at least three algorithms that have been developed: destination-based, source-based,

and XOR (refer to

Table 4-14

). The source-based algorithm utilizes the last one or two bits

(depending on the number of ports in the FEC/GEC group) of the source address in the packet.

If the bit is 0, the first port is used. If the bit is 1, the second port is used. The destination-based

algorithm utilizes the last one or two bits (depending on the number of ports in the FEC/GEC

group) of the destination address in the packet. If the bit is 0, the first port is used. If the bit is 1,

the second port is used. The XOR algorithm utilizes the last one or two bits (depending on the

number of ports in the FEC/GEC group) of the destination AND source addresses in the packet.

The algorithm “XORs” the bits. If the result is 0, then the first port is used. If the result is 1, then

the second port is used.

FEC/GEC has developed into not only a switch-to-switch technology but also a switch-to-node

technology. In most cases, the node is a Multi-homed server with network port drivers that

support FEC/GEC. Problems can arise with switches using the destination-based algorithm when

switch-to-node FEC/GEC is used. Because the destination address of the FEC/GEC node is always

the same, the switch always sends traffic to that server on the same port. Because of this, receive

traffic is not evenly distributed across all ports in the FEC/GEC group.

Table 4-13 Example of load-balancing algorithms #1

Preliminary information:

MAC address of a two port FEC/GEC

group

Last byte (01) conversion to binary


group



group


MAC address of client1












Packet 1 is a frame transmitted from client 2 to the two-port FEC/GEC group.

Packet 2 is a frame transmitted from client 1 to the two-port FEC/GEC group.

Table 4-14 Example of load-balancing algorithms #2

Method 1: DESTINATION-BASED ALGORITHM (method used by HP teaming)

frame is transmitted out port 2

Packet 1 - Destination MAC address: 00-00-00-00-00-01

Last binary bit = 0000 000

1

70

The Mechanics of Teaming for the Advanced User

Section	Page
Network Adapter Teaming White Paper	1
Table of Contents	3
1 Introduction	9
Abstract	9
How to Read This Document	9
Section Layout	9
2 An Executive Overview of Teaming	11
What is HP Integrity Network Adapter Teaming?	11
The Goals of HP Integrity Network Adapter Teaming	11
Fault Tolerance: Self-Healing Network Connectivity	11
Load Balancing: RAID 5 for Server Network Adapters	12
Why HP Integrity Network Adapter Teaming?	13
3 Teaming Fundamentals for the First-Time User	15
A Technical Overview of HP Integrity Network Adapter Teaming	15
Fault Tolerance: Dual Homing or Network Adapter Teaming?	15
Load Balancing: Server-Based Routing Protocol or NIC Teaming?	17
Overview of Team Types	17
Network Fault Tolerance Only (NFT)	18
Network Fault Tolerance Only with Preference Order	18
Transmit Load Balancing with Fault Tolerance (TLB)	19
Transmit Load Balancing with Fault Tolerance and Preference Order	19
Switch-assisted Load Balancing with Fault Tolerance (SLB)	19
802.3ad Dynamic with Fault Tolerance	19
Switch-assisted Dual Channel Load Balancing	19
802.3ad Dynamic Dual Channel Load Balancing	20
Automatic	20
How to Choose the Best Team Type	20
Basic Teaming Versus Advanced Teaming	22
Basic Deployment Steps for HP Integrity Network Adapter Teaming	23
Download	23
Install	23
Launch	23
Overview of the HP Network Configuration Utility	23
NCU Main Page	24
Primary Actions from the NCU Main Page	24
Team Properties Page	24
Team Properties Page Tabs	25
Adapter Properties Page	25
Adapter Properties Page Tabs	25
Getting Started	26
Example Deployment of Basic Teaming	26
Example Deployment of Advanced Teaming	26
4 The Mechanics of Teaming for the Advanced User	29
Section Objectives and Prerequisites	29
Architecture of HP Integrity Networking Adapter Teaming	29
The “Networking Layers” of an Operating System	29
Teaming Software Components	30
Network Adapter Miniport Driver	30
Teaming Intermediate Driver	30
HP Network Configuration Utility	30
HP Teaming and Layer 2 Versus Layer 3 Addresses	30
Types of HP Integrity Network Adapter Teams	31
Network Fault Tolerance Only (NFT)	31
Network Addressing and Communication with NFT	32
Scenario 4–A: A Device Pings an NFT Team on the Same Layer 2 Network	32
NFT Redundancy Mechanisms	34
Basic Redundancy Mechanisms	34
Advanced Redundancy Mechanisms	37
Active Path	37
General Description of Active Path Operation	38
Active Path Configuration	45
Fast Path	47
General Description of Fast Path Operation	51
Fast Path Configuration	52
Redundancy Mechanism Priority	54
Redundancy Mechanism Comparison Chart	54
Failover Events	55
NFT Network Adapter Failure Recovery	56
NFT Applications	56
Recommended Configurations for NFT Environments	57
Network Fault Tolerance Only with Preference Order	57
NFT with Preference Order Applications	57
Transmit Load Balancing with Fault Tolerance (TLB)	57
Network Addressing and Communication with TLB	58
Scenario 4–B: A Device Pings a TLB Team on the Same Layer 2 Network	58
Transmit Load Balancing Methods (Algorithms)	60
TLB Automatic Method	61
TLB TCP Connection Method	62
TLB Destination IP Address Method	62
Scenario 4-C: A TLB Team Using IP Address-Based Load Balancing	63
TLB Destination MAC Address Method	64
Scenario 4-D: A TLB Team Using Destination MAC Address Load-Balancing Method	65
TLB Round Robin Method for Outbound Load Balancing	66
TLB Redundancy Mechanisms	67
TLB Network Adapter Failure Recovery	67
TLB Applications	67
Recommended Configurations for TLB Environments	67
Transmit Load Balancing with Fault Tolerance and Preference Order	68
Transmit Load Balancing with Fault Tolerance and Preference Order Applications	68
Switch-Assisted Load Balancing with Fault Tolerance (SLB)	68
Network Addressing and Communication with SLB	69
SLB Outbound Load-Balancing Algorithms	69
SLB Inbound Load-Balancing Algorithm	69
Cisco EtherChannel	70
Switch Vendor Port Trunking Technology Supported by SLB	71
SLB Redundancy Mechanisms	72
SLB Network Adapter Failure Recovery	72
SLB Applications	72
Recommended Configurations for SLB Environments	72
802.3ad Dynamic with Fault Tolerance	73
Switch-assisted Dual Channel Load Balancing	73
Dual Channel Configuration	76
Dual Channel Transmit Balancing Algorithm	77
Dual Channel Receive Load-Balancing Algorithm	77
Receive Load Balancing on a Per-Port Trunk Basis	77
Receive Load Balancing on Multiple Trunks	77
Dual Channel Redundancy Mechanisms	78
Basic Redundancy Mechanisms	78
Advanced Redundancy Mechanisms	78
Dual Channel Network Adapter Failure Recovery	78
Dual Channel Applications	78
Recommended Configurations for Dual Channel Environments	78
802.3ad Dynamic Dual Channel Load Balancing	79
802.3ad Dynamic Dual Channel Configuration	79
Automatic	79
Team Types and Redundancy Mechanisms Interoperation	79
Team Type and Redundancy Mechanisms Compatibility Chart	79
Mechanism Priority	80
Team Status and Icons	81
Network Adapter Teamed Status	81
Team State	83
Team Icons	83
HP Integrity Network Adapter Teaming and Advanced Networking Features	84
Checksum Offloading	84
802.1p QoS Tagging	84
Large Send Offload (LSO)	84
Maximum Frame Size (Jumbo Frames)	84
802.1Q Virtual Local Area Networks (VLAN)	85
4–7 HP Integrity Network Adapter Teaming Configuration and Deployment with Integrity Essentials Rapid Deployment Pack (RDP)	86
5 Teaming Feature Matrix	87
A – Overview of Network Addressing and Communication	89
Layer 2 Versus Layer 3	89
Address – Unicast Versus Broadcast Versus Multicast	89
Example Scenarios of Network Addressing and Communication	89
Scenario A-1: One Device Pings Another on the Same Layer 2 Network	90
Scenario A-2: One Device Pings Another on a Different Layer 2 Network	91
B – Frequently Asked Questions	95
C – Overview of Utilities Included with HP Integrity Network Adapter Teaming	99

HP Integrity rx2800 HP Integrity Network Adapter Teaming Whitepaper - Page 70

Cisco EtherChannel

Page 70 highlights