9

Doing my first server to iSCSI target and I thought i would be off to a running start by teaming my NIC on the server (conventional 802.3ad team).

However, from many online sources I see this is NOT recommended and MPIO should be used instead (there was no specific sizing in this recommendation and seemed to apply from a 3 node cluster to large clusters of tens of nodes and above).

I know MPIO is better as it aggregates transfers over independent routes (whereas conventional NIC team would be to the same switch), this aside is there any other reason (risk of data corruption) for Microsoft and training material state to use MPIO instead of teaming?

1 Answer 1

11

Can you expand a bit on your iSCSI architecture? How many initiator/target addresses are you working with, how many physical switches, all one subnet or multiple?

The basic answer is: because MPIO manages end-to-end connectivity paths, and is better at storage connectivity load balancing and connection resilience than generic network redundancy and load balancing mechanisms.

The specific technical reasons for this depend on the architecture, so I can be a lot more specific if you provide additional detail on your iSCSI network's setup. A few general examples:

  • Without any MPIO, your initiator-to-target IP conversation is just a single conversation. 802.3ad mandates that the order of packets in a conversation not be changed (and you wouldn't want your iSCSI traffic out-of-order anyway), so you're limited to the bandwidth of a single link.
  • MPIO detects and handles path failures, whereas 802.3ad can only compensate for a link failure - and only if that link failure is correctly detected. If your NIC card hangs but still reports good link, or your switch configuration gets screwed up for a specific port, you will likely lose storage connectivity despite having a second link that's still working.
  • You're tied to a single physical switch, instead of being able to uplink your host's NICs to different switches.
3
  • Hi Shane, thanks for the reply - for our setup it is a single storage subnet and 1 physical switch, however the documentation didnt state a size it just said to use MPIO only. From your comments it seems it is certainly possible to use teaming, but MPIO is preferable. Good explanation, unless you want to add anymore you have answered my question and will mark it as such soon.
    – morleyc
    May 25, 2013 at 20:03
  • @g18c Yup - so the main advantages you'll get from MPIO instead of link aggregation would be the potential for higher throughput (if your throughput is enough to saturate a single NIC), as well as being able to gracefully handle a wider range of failure states. I guess the question is, are there any particular reasons that you're preferring to use 802.3ad instead of MPIO? May 25, 2013 at 20:08
  • MPIO gets particualrly usefull when you ISCSI setup is more advnaced - because with MPIO ou have separate logical connnections which can go do redundant endpoints. The SAN can replicate in the background and the 2-3 targets actually are different amchines. As long as you have a "single path, single NIC" scenario, the advantages are not that brutal.
    – TomTom
    May 26, 2013 at 11:15

You must log in to answer this question.

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