VMworld 2014 Roundup: Day 4

There was no general session on Wednesday.  I managed to take in a session, sat in as the audience for the live VMTN Community Roundtable podcast and spent a good few hours in the Solutions Exchange.

NET2745 – vSphere Distributed Switch Technical Deep Dive

Delivered by Chris Wahl (@ChrisWahl) and Jason Nash (@TheJasonNash), this session provided information, some recommended practices and discussion on the vSphere Distributed Switch (vDS). In memory of Robin Williams, let’s try this stream-of-consciousness style.

The vDS control plane is contained with vCenter. The data plane within the ESXi hosts. Use elastic ports! Connect your uplinks to a single network segment. Use 802.1q tags for port groups. Make sure you have at least two (2) uplinks per vDS. A 2 x 10GbE config can work fine. Enable QoS tagging on your vDS or physical switches, but not both. It’s recommended to use descriptive naming everywhere as you build out your networking config, such as including the VLAN ID in your port group name.

At this point in the session Chris and Jason pull off a real-time demonstration of standard vSwitch to vDS migration. Much to the delight of the audience the demo works without a hitch, despite the last minute replacement of Transformer object names with SpongeBob names.

There’s a bit of a discussion about dealing with a mix of 1GbE and 10GbE interfaces, for example while transitioning from 1 to 10 GbE. The recommended approach is to leave your 1GbE VMNICs connected to their original port group and create a new port group to connect your 10GbE VMNICs to.  Move your management and vMotion traffic from the 1GbE PG to the 10GbE PG.

A quick discussion on how to control vSphere Replication traffic by establishing traffic limits. This way you don’t saturate your link during peak usage periods. The default UIs leave a bit to be desired as the admin has to manually change the limits. Chris Wahl has put together a PowerCLI script that will all you to control the traffic limit value. This is especially helpful if you want to schedule a limited amount of replication traffic during business hours and then open it up overnight.

One last tip is to look into load based teaming to make sure there is better load distribution via balancing. For example consider leveraging the “Route based on physical NIC load” teaming policy for your vDS to implement “truer” load balancing. Make sure the extra processing overhead doesn’t cause any issues.

This was a great session from Chris and Jason and I think we can contribute the majority of its success to the fact that Jason never put his coffee cup down the entire time. Well done!

VMTN Communities Roundtable Podcast c/o #vBrownBag

Eric Nielsen (@ericni25) and Mike Laverick (@Mike_Laverick) hosted this week’s VMTN Communities Roundtable podcast on the #vBrownBag Community Stage at the VMworld Hang Space. The episode was recorded via TalkShoe as it normally is and was also live streamed on the vBrownBag UStream.

Eric and Mike summarized and chatted about the announcements made thus far at VMworld. For example VVOLs, vSphere 5.8, vSphere 6.0 beta and of course EVO:RAIL and EVO:RACK.

Check out the show notes for a link to the recording as well as references for the topics discussed.

Solutions Exchange

I managed to chat with a few vendors in the Solutions Exchange, including an interesting look at CloudPhysics. I’ll be providing more details in the upcoming Vendor Spotlight posts.

Wrap ‘Er Up

Just one day’s worth of Roundup left to round out our VMworld summary post series. Stick around.