There are a ton of what’s new posts coming out this week about vSphere 5.1. Many of them probably covering the overall new features but I wanted to focus primarily on networking, as that’s a piece I work with very often and what I’ll be talking about at VMworld 2012. Bottom line is this, VMware has done a great job with the VDS (notice it’s now all cap’d) in 5.1. They’ve added features that were often requested and added safety nets for those that weren’t 100% comfortable running everything through the distributed switch.
First, the bullet list of new features:
LACP has been requested by many since the release of the VDS..and honestly, before that with just standard vSwitches. It was finally added to the Uplink configuration of the VDS so now you can use LACP for negotiation of your port-channels. About time…and glad it’s finally here.
ERSPAN and RSPAN are enhancements to the port mirroring capability we got in the vSphere 5 VDS. Now you can encapsulate and do remote span, just like we’ve been able to do with the Nexus 1000v. One nice addition here too, is the improved workflow to configure port mirroring sessions. It’s a lot more intuitive and user friendly than in the past.
Configuration backup and restore is just like it sounds. You can now backup your VDS configuration to a file and then later restore it should something happen. This is one of the new safety nets added to the VDS.
Rollback and recovery will help to convince people that it’s okay to run everything through the VDS. It’s cool. Let’s say you’re on the vSphere 5.0 VDS and you migrate your management vmkernel interface to the VDS, but you didn’t set VLAN tagging correctly on the VDS side. What happens? Your host drops and you have to go (usually) to the DCUI and recover. Not anymore! After a 30 second timeout if vCenter and the vSphere host can’t communicate your last change is rolled back and an alert is thrown. It recovers the host for you. Great, great new feature.
Health check is an interesting one. It confirms VLAN trunking and MTU size between the VDS and the upstream physical switches. This requires a minimum of two physical NICs in the uplink configuration for the switch. If you set your VDS frame MTU size to something like 9000, but the physical switch is set to the usual 1500 you’ll get an alert. You’ll get a similar alert if you create a port-group to use VLAN 394 but VLAN 394 isn’t being trunked from the physical switch. This check is set to run automatically every so often so you turn it on and basically forget it until you get an alert.
NetFlow v10 is now supported, but not NewFlow v9.
Better configuration and management workflows across the switch. I already mentioned the new port mirroring workflow but there are more. Another example is host migration to the VDS. They’ve reworked this in a very good way. You can now go back and use a wizard to migrate the rest of a host’s networking to the VDS if you only did a simple add to start. You can even migrate the rest of some hosts networking and add new hosts at the same time, so that the configuration is standard. It’s much better than it used to be.
Dynamic port allocation simply means you no longer have to tell the switch how many VDS ports you want for VMs. It will start low and automatically add as needed.
SR-IOV (Single Root I/O Virtualization) is now supported on the Enterprise Plus version of vSphere 5. This allows a PCIe device, that supports SR-IOV, to appear as multiple PCIe devices. We do this a lot with things like Cisco’s VICs in their UCS servers. It allows you to “divide up” something like a NIC in to a bunch of virtual NICs where you can apply policies to them, or pass them through direct to a VM. The downside to SR-IOV is that it does keep you from doing things like vMotioning a VM.
There are many other minor changes, performance enhancements, and tweaks as well. Really great job VMware.
Here is a video where I walk through many of the new features: