Private networks that are enabled for cluster use get a metric starting at 1,000, and each additional private network get a metric incremented by 100. So the first private network enabled for cluster use gets a metric of 1,000, the next gets 1,100 and so on. If a private network isn't configured for cluster use, then it will get a much higher metric than public networks that are available for cluster use. Public networks that are enabled for cluster use get a metric starting at 10,000 and increment by 100 if there are multiple networks.
The cluster fault tolerant driver (NetFT.sys) uses the network with the lowest metric for the cluster Internet traffic, such as heartbeats, and for Cluster Shared Volume (CSV) traffic, which is why you need a 1 Gbps or 10 Gbps network for a cluster network if you're using CSV.
For the network used by live migration traffic, which is the memory of the virtual, delta copies of the virtual and then partition state (as well as CPU/device state, but this is a tiny amount of traffic) each virtual machine (VM) can have its own network priority order. By default, the nework with the second lowest metric is used while the lowest metric network, the one used by NetFT, is actually placed at the bottom of the network priority. This is because you don't want CSV traffic and live migration traffic going over the same network.
Source: http://windowsitpro.com/mobile/pda/Article.cfm?ArticleID=101857&FAQ=1
No comments:
Post a Comment