Trunking on a Gigabit Network Interface

[hp_user_1]

Hello,

Can someone help on how to divide a single gigabit interface into 3 logical interfaces with each having a separate IP address and hostname in HP-UX 11iv3?

Thanks

[Mel Burslan · Accepted solution]

in the /etc/rc.config.d/vlanconf file, you will need to have 3 of these blocks:

VLAN_PHY_INTERFACE[0]=lan0
VLAN_ID[0]=701
VLAN_PRIORITY[0]=0
VLAN_TOS[0]=0
VLAN_PRI_OVERRIDE[0]=CONF_PRI
VLAN_TOS_OVERRIDE[0]=IP_HEADER
VLAN_NAME[0]=SAPTHRUCSS
VLAN_VPPA[0]=5000


the first line refers to the physical interface of your own selection, the VLAN_ID should be negotiated between you and the guy in charge of network switches so that bot sides tag the frames with the same ID. And the last line VLAN_VPPA should be unique for each virtual interface. Also each of these 3 blocks defined, should have a different index number between the square brackets, i.e. [0], [1] and [2].

Then you need to define the ip addresses for these VLAN_VPPAs in your /etc/rc.config.d/netconf and their name will show like lan5000 for the above example.

Hope this helps

[hp_user_1]

Thanks Mel,

Is it possible to configure bandwidth for each of these virtual interfaces ? Say 200mb for IP 1, 300mb for IP 2 and 500mb for IP 3?

[Mel Burslan]

I do not believe you can control the bandwidth on the server side. It is more of a function of the network gear as far as I know, but again, I might be mistaken. We never did this. Again just by looking at the definitions of the lines in vlanconf file, it seems like you can prioritize the vlan interfaces with VLAN_PRIORITY line(s). So giving the higher priority to your higher bandwidth requiring interface might be the solution, although this does not directly map onto limiting bandwidth.

[rick jones]

VLANs are indeed one way to divy-up a single interface. HP-UX also supports the concept of a "logical" interface that does not use VLAN tagging. If you had lan0 you would then assign it one of the three IPs and then add a lan0:1 and lan0:2 interface with the desired IPs to your /etc/rc.config.d/netconf file - via Sam or the SMH or whatever it is 11iv3 wants to use these days.

As for bandwidth constraints on each of those, it might or might not still work but there is a now very old port of "dummynet" on ftp.cup.hp.com under dist/networking which if you did go the VLAN route could be pushed onto each vlan interface. it would have to be "enhanced" to work with three separate logical interfaces - those are just one DLPI instance where IIRC the VLANs are separate instances as seen by the transport.

There may be enhancements to ipfilter, but that is just a wild guess.

Over in the ProLiant space, there is the recently announce "Flex-10" functionality in ProLiant Blades. I cannot say when/if said functionality would be available in an Integrity Blade, but I suppose you could ask an HP sales rep. In its current form, Flex-10 NICs are dual-port 10G Ethernet interfaces where each port can be subdivided into as many as four "flex-nics" and assigned a bandwidth in 100 Mbit/s increments. That is handled entirely by the NIC/blade I/O module without additional overhead in the host.