Direct-connect LACP

[petr]

Hi,

I've read your primer about LACP and it seems to me that I should be able to use round-robin packet distribution, my setup is following:

- FreeNAS connected to the main network via an onboard ethernet port
- FreeNAS with 4port (1Gbps) expansion card
- Mac Pro with 1 ethernet connected to the main network
- Other ethernet port connected directly to one of the ports on the card
- 2x additional Thunderbolt ethernet adapters directly connected to other ports on the expansion card of the NAS

I've got a few Cat6 cables running between the FreeNAS box and my main workstation and I would like to improve performance for transferring large files.

My plan is to keep using the FreeNAS via the onboard LAN port with all other clients but I would like to create dedicated bonded line out of 3x1Gbps direct links.

From my reading it seems that this may work - any thoughts on this? Would AFP give me full available bandwidth?

 

[cyberjock]

If you are hoping to get more than 1Gbs/sec from your one workstation to your server, you HAVE to buy 10Gb networking gear. LACP helps with many clients and a single server. You cannot exceed 1Gb/sec with LACP if your lan is only Gb.

If what you are trying to explain is what I think you are, you cannot bond 3x1Gb NICs between your server and client machine and expect 3Gb/sec. It doesn't work that way. The LACP primer explains this. :P

 

[petr]

Hmm I must have misunderstood this:

3d) It is possible to instruct FreeBSD to use "roundrobin" hashing, to accomplish what most end users seem to expect LACP to do. Unfortunately, this ultimately results in violations of the misordering prohibition noted in 3a), and while it often appears to work swimmingly well on a lightly loaded network, it typically devolves into rapid hell as the network gets busy, because packet queuing delays become a major factor - just the thing you DON'T want to happen.

From the above I thought it was indeed possible to coerce FreeNAS to do round-robin distribution, meaning that the interfaces' bandwidth could be added up...

My understanding is that it is not possible by default with the hash-per-stream based papproach, however, wouldn't the round-robin effectively utilise the bandwidth of all interfaces for a single stream?

 

[cyberjock]

Round Robin doesn't add up bandwidth. It only ensure that your traffic is evenly distributed across the lagg. Nothing more and nothing less.

In a realworld example, it prevents you from having a bunch of clients bottlenecked by one nic while the other nic is not busy.

You will still be bottlenecked at 1Gb/sec.