Are you guys out there uses virtual switches? If you are, are you using the Cisco 1000v or VMWare vSwitch or something else? What challenges do you face in monitoring and managing virtual switches?
currently virtual switches, planning to go to nexus.
currently no monitoring at all for these (dont know how)
What types of questions would you like answered? What problems do you face with these switches? Example, a user cannot see the traffic that goes back forth between hosts on the same vSwitch on the same ESX box.
We're currently using the VMWare vSwitch but have considered the 1000v. I would love to get more detail about the networking side of things. Right now, we monitor the NICs on the ESX servers and then the NICs on the hosts. We get a little information but we have one Blade Chassis that would be great to monitor in more detail.
Our server team is moving away from HP blades and installing Dell servers running VMWare. The first set of these new servers are connected to Cisco 5ks and 2ks.
Soon we will be installing 1000v. Previously our biggest concern was limited view into the "vmware" world. Hopefully when the 1000v comes online, we will have better visibility into that environment.
I have not researched the NPM capabilities for the 1000v, so will we have a better view inside this virtual world?
We have looked at the v1000 but the same problems exist when monitoring any virtual switch.
How much CPU and RAM will monitoring eat up? Since the vSwitch is nothing more than a piece of software running on a server the VMware host's resources will be used to produce the stats.
Bottom line: A baseline is needed to guesstimate how will monitoring virtual switch stats drag down the performance of the VMware host.
Using the vSwitch right now. Have looked at the 1000v. There is little visibility into the vSwitch. 1000v seems to solve it at a cost (money & server resources).
Right now there is no easy way to tell which pnic a guest is using (we are using etherchannel trunks). Nor is there a way to see internal traffic or stats on each virtual port.
I know we've looked at the 1000v but I haven't heard anyone say we're going to them. We currently have no visibility into the vSwitches on the ESX servers. You're right, we can't see any host to host traffic for servers on the same ESX box. We use Opnet products for packet analysis and they make a virtual version of their packet capture appliance we've considered. However, it's not an issue high on anyone's list right now with budgets being what they are.
Good discussion. Let me bubble up the points here
