Shulou(Shulou.com)06/01 Report--

1. What is Teaming?

To put it simply, Teaming is to bind multiple physical network cards on the same server into a virtual network card through software, that is to say, for the external network, this server has only one visible network card. For any application, and the network on which this server is located, this server has only one network link or only one accessible IP address. The reason why we want to use Teaming technology, in addition to using multiple network cards to work at the same time to improve network speed, we can also achieve load balancing (Load balancing) and network card redundancy (Fault tolerance) among different network cards through Teaming.

2.Teaming species

2.1.Broadcom Teaming

There are three ways to implement it:

 Smart Load Balancing

 Link Aggregation (802.3ad)

 Generic Link Aggregation (Trunking)

2.1.1.Smart Load Balancing (SLB)

SLB can achieve load balancing and network card redundancy, and is independent of the switch model (applicable to most brands of switches). Specifically, all physical network cards on the server have separate MAC addresses, and the incoming and outgoing data packets are controlled by the SLB program. Once the control program detects that one physical network card is damaged, it will automatically distribute the traffic to other working network cards. This ensures the continuity of the network work, and the program that accesses the server through the network is transparent to the outside.

In addition, in this mode, different physical network cards can be connected to different switches, thus achieving switch redundancy.

The following is the support of different operating systems for load balancing in this mode

The following is the support of different operating systems for network card redundancy in this mode

2.1.2.Link Aggregation (802.3ad)

This mode is controlled by the Link Aggregation Control Protocol (LACP) protocol and is divided into dynamic and static configuration modes (set on the switch). In this mode, virtual network links receive packets using only one Mac address (the Mac address of the first physical network card in the Teaming).

In addition, in this mode, the Teaming program is only responsible for load balancing the outgoing packets, while the incoming packets are balanced by the associated switch, so the switch connected to the server is required to support the IEEE 802.3ad standard. The Teaming program and the switch are jointly responsible for monitoring the link status and automatically switching if a link is found to be defective.

The following is the support of different operating systems for load balancing in this mode

The following is the support of different operating systems for network card redundancy in this mode

2.1.3.Generic Link Aggregation (Trunking)

This model is an extension of the second model, and different companies localize Link Aggregation, such as Cisco's Fast EtherChannel (FEC), Cisco's Gigabit EtherChannel (GEC), which are all Trunk protocols. In the application, the administrator must statically configure some ports on the switch, and here the Teaming program and the switch jointly monitor the link status.

The following is the support of different operating systems for load balancing in this mode

The following is the support of different operating systems for network card redundancy in this mode

2.1.4. Comparison of three modes

See the following list for details

Schematic of 2.1.5.Teaming Network Topology

Because of the danger of data circulation, the above topology method 1j2 requires that the corresponding protocols to prevent data circulation, such as STP protocol (for cisco switches), should be opened on the switch.

2.2.Intel Nic teaming classification

The classification of Intel network cards consists of five categories.

Redundancy of  network card

 switch redundancy

 network card load balancing

 Link Aggregation (FEC)

 Link Aggregation (GEC)

 IEEE 802.3ad

2.2.1. Nic redundant Adapter Fault Tolerance (AFT)

In this way, if there is a problem with a network card, other network cards in the Teaming will be automatically topped up to support 2-8 network cards without load balancing function. in addition, this mode has no special restrictions on the switch, but only requires that it be connected to the same switch.

2.2.2. Switch redundant Switch Fault Tolerance (SFT)

Two network cards are supported to be linked to different switches for link redundancy. Note here that the switch needs to start spanning Tree Protocol (STP) to prevent loops, which has no restrictions on the switch.

SFT only runs on Windows NT 4.0, Windows 2000, and Windows Server 2003

2.2.3. Load balancer Adaptive Load Balancing (ALB)

This mode realizes load balancing and redundancy. Under the windows operating system, it can also control the teaming in real time, such as turning off the load balancing of incoming data. There is no restriction on the switch in this mode.

2.2.4.Fast EtherChannel*/Link Aggregation (FEC)

This mode provides 8 network cards of 2Mel, which can be linked to the switch at a speed of 100m for load balancing and redundancy, but requires the switch to support the corresponding protocols, such as Cisco switch.

2.2.5.Link Aggregation (GEC)

This mode is an extension of FEC, but does not work at 1000m, and the switch port needs to be configured with static mode (static).

2.2.6.IEEE 802.3ad

In this mode, the network card can work at different speeds, that is, the teaming can be established with the network card that cannot get through the speed, but the switch is also required to fully support the IEEE 802.3ad standard.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.