0

I was wondering something related a setup I was planning to do.

I am redo doing my small vmware network to change to a better box as my iscsi target. Just like my old box this one has 2 gigabit network interfaces. My ESXi box also has 2 interfaces that I dedicate to for storage.

The problem is I don't have a switch that supports jumbo frames and load balancing. Its just a dumb gigabit switch.

My question is, if I connect the 2 interfaces from my iscsi target into the 2 interface on my ESXi box will they still load balance over both interfaces? Or because there is no switch the interfaces will just send from one end to the other?

I hope this makes sense, been up for a long time not all here.

Thanks in advance.

Improve this question

3 Answers 3

Reset to default
1

iSCSI should never use bonded interfaces because of the way it works. Use MPIO for iSCSI redundancy and increased throughput.

Improve this answer
4
  • You wouldn't happen to have a reference in the docs for that? I'm about to setup a test environment and would like to learn a little bit more about it.
    – Justin
    Apr 14, 2011 at 19:16
  • @Justin, first thing I found: forum.synology.com/wiki/index.php/…
    – Chris S
    Apr 14, 2011 at 19:19
  • 1
    I was in the middle of typing up a long-winded thing that you've covered much better with much less verbage. @Solignis, your old setup wasn't really load-balanced; the VMware load balancing modes on the vSwitches that don't require switch support route based on either source MAC or source virtual port ID; everything will hit one NIC for the storage traffic because it all comes from the vmkernel interface.
    – Shane Madden
    Apr 14, 2011 at 19:21
  • Very very cool, I will have to see if I can get this to work for me. I did not think you could do MPIO on ESXi without a paid version. Seems to give me the option to do it without a paid license. Thanks for the help.
    – AtomicPorkchop
    Apr 15, 2011 at 13:38
0

iSCSI w/o Jumbo frame on a dumb switch...I'd start by upgrading to a Managed switch with Jumbo Frame. That should boost the performance more than Nic teaming IMO.

Also, Link Aggregation in ESXi is done with IP hash by default. So if both source and destination are always the same, it will never load balance.

Improve this answer
0

I will echo what has already been said, but using link aggregation in some scenarios is not a bad idea. If for example you have Linux clients with multiple interfaces on redundant switches, as in perhaps 2 interfaces on one switch 2 interfaces on another switch, you can build your bonds to increase bandwidth up to your switches, and of course, I am assuming that your iSCSI system will also have multiple paths, direct/indirect to the same switches. You can leverage multipath on Linux and configure round-robin, assuming you also define which bond will tie back to which IP of your iSCSI target. There are many reasons to do it, though I agree with others, with ESX/ESXi use VMWare's MPIO stack.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .