Has anyone been able to switch a K2000 VM to use a 10gig adapter?
I know we don't have any root access, can't log in as a sudoer. I know the K2000 reports that it's got VMWare tools installed so the proper drivers should be installed. I tried configuring one of my RSA machines with a vmxnet2 adapter and it was no-go. If I could use lspci, etc. as a sudoer I think it'd be pretty easy to get the 10gig adapter running. I hope they eventually start thinking of the VM's as a software package instead of a hardware package and give us a little more access :grrrrr:. Any thoughts? I'd love to actually be able to utilize the network bandwidth I've got available.
Answers (2)
I was able to port team to the VM to do a 2 GB throughput. I just present two network adapters for the same network in the hardware for the virtual machine, and then port team them in the switch. Try going at it like this from the network / VM side. I may be getting away with it because I'm using two 10/100/1000 pipes, and the K2K understands a 10/100/1000 NIC, but not a 10G NIC. You may be able to present ten 10/100/1000 NICs and consolidate them later with port teaming back at the switch.
Comments:
-
If it helps, I agree, I'd like more control over the internals. I understand that KACE likely keeps us out so they don't have to support the modifications we may make to the OS. Right now, the OS is generally a known value that they can handle. If they allow us in, they have to deal with that as a possible reason the KACE is failing when we ask for support, and both sides have to deal with it when deploying upgrades. That said, they do need to create ways to adjust things appropriately with controlled access to needed functionality, such as resizing disk, changing network configurations, etc. - philologist 12 years ago
-
philologist - did you actually verify you're getting bonded throughput? from my work with some of the KACE engineers they said that only one of the ports actually transfers data when imaging. Basically, when both ports are configured one's doing imaging and the other's management. Thoughts? - lcadhelpdesk 12 years ago
-
I can say I saw much better throughput, but that's about the only thing we can look for; I don't have access to the OS internals to verify transfer speeds. To my knowledge, I haven't given it two separate IPs, so it would have to either be using load balancing, or QoS internally to segregate traffic like that. I would think I'd be seeing all kinds of problems if it isn't accepting one of the virtual adapters and is just dropping the packets, but again, I have no access to the OS. - philologist 12 years ago
-
gotcha. from what I've observed by checking port activity, etc. it's still only using one interface. The secondary interface is for external storage (at least on 3.4). Unfortunate. If I had more access to the box I could add more adapters :rolleyes: - lcadhelpdesk 12 years ago
-
If it means anything, our helpdesk tends to order machines 25 at a time. During testing, I did not have a problem deploying to all 25 at once. Our image takes about an hour to push from the K2000 before it gets to post-install tasks and sysprep where the K2000 is no longer part of the equation. With 25 laptops (slow drive access, which I expect is the real bottleneck in the system), this slowed down to about two hours at worst. I expect it would take quite a few machines to get to the point where they weren't complete if you started the push and came back in the morning. - philologist 12 years ago
VK-series appliances (just as any guest OS) utilize the network at physical speeds allowed by the hypervisor. Actual throughput it determined by the physical infrastructure. For example, if an ESXi host has 10GB connectivity to the network, a VK2000 can use that bandwidth. If using VMware vCenter, the VK2000's Performance tab contains network utilization information to monitor actual network utilization. There's no need to change the E1000 NIC driver in the guest OS to utilize maximum host bandwidth.
