Hello,
I'm a bit confused when trying to come up with an estimated maximum speed on my iscsi network.
Let's say I have 1 esxi host with 2 nics (1gb) in a single vswitch with their own vmkernel where only one adapter is active and the other one is unused. Then I have 1 san with 3x1gb iscsi configured nics.
I test with crystaldiskmark and in sequential reads and writes I get something close to 220 MB/s which makes sense for me since I cannot possibly be faster than 2x1gb nics (250MB/s would be the max).
Then I say to myself, if I add 1 or more iscsi nics, I could get close to 330MB/s since then I would have at least 3 nics on the host talking to 3 nics on the san. When I run the same test again, I get almost the same numbers...
So then I start to read a lot on the subject and I find that 1 vm can only initiate 1 iscsi session which can only use 1 physical nic at the time when using the teaming policy route by port id...
I scratch my head then asking, how can I achieve 220MB/s then if I can use only 1 physical nic ?
To make a long story short, I'm puzzled as to how I can exceed the maximum speed of a physical nic (if it really works that way) and if it doesn't, why do I get almost exactly the same result when adding an additional iscsi nic.
I know there's the whole raid type, nb of spindle, etc... But I'm focusing on the network part of the equation right now and I cannot predict the behavior very well since I don't seem to understand the inner working of the iscsi network.
I'm using the round-robin policy with iops=1, jumbo frames working all across the iscsi network also.
Please help me to understand.
I don't know if I'm posting this at the right place, if not please tell me where.
Thanks.
