ISCSI SAN configuration questions

Great information Mastoo.  I do not have a good assesment of our i/o, but I did send over some perfmon results for EMC, at which point they never gave us back the kind of exact metrics that we were hoping to get... I called them up a few weeks later and simply got a "you will be fine" answer.  

I assume that when you are speaking of the active/passive you are referring to the procs and the passive essentially being a backup... is this correct?  (I am very much a beginner in the SAN world)

So, it is sounding as though I will only ever have a single cable connection at any point, therefore my question is this; assuming that we stick with this setup and dont look at returning it, and assuming that we detirmine we can live on 80MB throughput, it really doesnt make since to team 4, 1GB NIC's on the server side to the switch... do you agree?  I might as well break the teams and offer up more single cable connections to the different hosts....

Thanks!

I was a complete newbie just 3 months back - the usual baptism by fire.  Yes, my active/passive was referring to a controller for redundancy.

Since we're doing ESX and you're Hyper-V the teaming thing might be different.  In ESX we set up multipath i/o on the host actually at the iScsi level, so I'm not sure we're talking the same thing.  Are you actually referring to teaming or bonding at the nic level?  I don't recall any EMC people mentioning that as an option when I was starting to sweat the bandwidth limitations.  We set the mpio to use a round robin as a means of distributing the i/o across all of the wires.  The possible benefits are: 1) A single virtual machine connection can get throughput spread across multiple wires; 2) The vm's rotate through the paths so they kind of are load balanced and less likely to step on each other; 3) Redundancy.  I think for your question #1 doesn't apply.  #2 and/or #3 might still justify some kind of pooling of the connections if that is how you accomplish that in Hyper-V.

We just plunged in with ESX so I never even looked at Hyper-V

I do think that we are running somewhat of a different configuration which may be causing a little confusion as well.  See in our environment, while we could present the SAN Luns to the VM's via the Hyper-V host, we have chosen to instead install ISCSI initiators on the VM's themselves, as this allows us to easily move our VM's from one host to another...  That said, the Hyper-V host that is running the hypervisor is not essentially aware of the SAN even.

When I refer to the teaming/binding, yes I mean at the NIC level... the original assumptions was to take say 2 ports and bind them to get a 2GB connection over to our swtich that goes into the SAN, and then present this teamed NIC to a VM as a dedicated connection to the SAN.  This was how we hoped to ensure that we did not have ANY extra network traffic interfering with our main SQL database.  

So in this situation, we would have our main SQL box with a normal network connection to the rest of our network and then would have a team of 4 NICs acting as 1 (giving it a 4GB connection) over to the SAN.... This team of 4 would not be presented to any other VM, therefore there would be nothing for the hypervisor to load balance.  This same setup would translate over to a few other machines.

I guess there are several ways to skin this, but the main thing that I am finding in reference to my original question is that it sounds like it can be done on some models but not the AX4.... I have a conf call in a bit with someone at EMC and Ill update with any other info.