Virtual Combat: The four teaming of network design

Network teaming this concept in the physical server has long been very common, we tend to set up a physical server a number of physical network card teaming, in addition to the protection of the network card caused by a single point of failure, there are load balancing purposes.

In a virtual environment, there is no need to connect multiple virtual NICs to a virtual machine for fault tolerance or load balancing purposes in most cases. Because fault tolerance or load balancing tasks are delivered to the virtual switch and the multiple physical network adapters it connects to. How to achieve it? This requires setting up the Networkteaming Policy on the set virtual switch.

Which of the five strategic choices will suit your environment?

Route based on originating virtual port

Route based on IP hashes (only one supported Withstatic EtherChannel and Static 802.3ad)

Route based on Source MAC address

Route based on physical NIC load (load basedteaming or LBT)

Use explicit failover order (not a load balancingalgorithm)

In addition to route based on IP hashes that require link aggregation to be set on physical switches, other policies do not require special settings on physical switches.

Scenario One:

A small company has not purchased the Enterprise plus license because of cost considerations, so use the vsphere Standard Switch.

Suggestions:

Using route based on originating virtual port

Depending on the port ID of the VM connected to the vswitch, it is quick and easy to transfer the packet to the corresponding physical network card without vmkernel any redundant processing of the packet.

See more highlights of this column: http://www.bianceng.cnhttp://www.bianceng.cn/Servers/virtualization/

Scenario Two:

A large company has many people, using vsphere distributed Switch. The ESXi connects six 1G physical network cards. There are several virtual machines as the company's file server, often employees complained that access to the file server delay serious, sometimes a file for a long time can not open.

Suggestions:

The route based on IP Hash is used (only onesupported with static EtherChannel and static 802.3ad). However, the prerequisite is that the physical switch must support Staticetherchannel or static 802.3AD link aggregation and set accordingly.

This strategy maximizes the throughput of the file server, because EtherChannel can bind multiple physical network cards to a channel, then throughput is changed from 1G to 1g* the number of network adapters within channel

The above example is a special case where multiple clients initiate file access to the file server virtual machine, and the throughput of the 1G network card does not meet the requirements.

If your application is in most cases point-to-point communication, which means that only one physical NIC can be used at the same time, routebased on IP Hash does not offer special benefits compared to LBT.

Scenario Three:

A company uses vsphere distributed Switch. The ESXi connects two 10G physical network cards. Because of the use of LVS or hardware devices to achieve load balancing, the current virtual environment does not require a particularly high network throughput of virtual machines.

Suggestions:

Using route based on physical NIC load (load based teaming ORLBT)

vSphere4.1 later, VDS supports this policy. When the connection is over 75% for a physical network card on the ESXi, the packet is automatically sent to other more idle physical network cards to achieve load balancing purposes. Note that although you set up multiple physical network adapters, at some point the packets are only passing through a specific physical network card. 10G is the maximum throughput that the VM can achieve.

If you use VDS, LBT is the best choice in most cases. There is no need for complex configuration of physical switches in scenario two.

This article comes from the "Sit Watch Cloud up" blog, please be sure to keep this source http://frankfan.blog.51cto.com/6402282/1227531