Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Arnow]

I'm interested by the results of your tests
Thanx

[JavaLawyer]

Have you had an opportunity to perform your link aggregation testing? 

[mrveronn]

Well, its a mit confusing. Wish the "Dummy's Guide to Ethernet Bonding" was available. 

I have the two LAN cables connected, and i've tried various settings, but the results are not consistent. I've seen network utilization rise...originally was 9% with the single connection and my Seagate drive. The transfer was fairly constant (+/- a few percent)...now ranges from 3% up to 60% (on a gigabit router/switch) in wild swings (looks like the squiggles from a lie detector tape)...!!

I've looked at the resource monitor available from the NAS' ... it shows graphs for LAN1 LAN2, but only LAN1 shows any activity...LAN2 never shows a twitch...regardless of the setting.  Am i doing it right?? What is the process for the DNS-345...its just not really clear. I've restarted the unit (as another post suggested), but noting every registers on the LAN2 graph.

I don't know how to approach this in a methodical manner...so have shelved it for now...left the settings on Adaptive Load Balancing...

[JavaLawyer]

I would imagine the dual ports would also be beneficial in a WAN environment, where two routers in the same WAN have dedicated GB connections to the DNS-345.

[JavaLawyer]

Wouldn't you need a managed switch (with bandwidth aggregation support) behind the DNS-345 to route inbound traffic? I'm looking at the DNS-345 manual and found this excerpt:

". . .The ShareCenter will auto-negotiate the highest connection speed available to your router or switch. If you are using Port Bonding, use a managed switch."

[mrveronn]

Well, that is part of the complexity.  Some of the options in the "link aggregation" require a switch that supports 802.3ad, which my D-Link router does not, but there are a number of options and some options do not need the specialized support.

I'm still reading on this, but here is an example of where someone was looking for performance gain, and decided that 802.3ad was not the way to go...

From my understanding the only bonding mode that offers any performance improvement for a single TCP stream is balance-rr (mode 0 for the bonding driver) which simply tosses packets out in a round-robin fashion.

...rather the balance-rr (Round Robin) was the 'performance choice' after his analysis. The article goes on to describe the challenges he faced (and overcame)...how he figured out how to make it work...and how to measure the throughput: http://devnerd.net/2011/02/high-performance-interface-bonding-in-linux/

But MTU (?) and most of this is gobble-de-gook to me...

[mrveronn]

According to this document, the round robin needs a switch or router that can be configured for "static link aggregation" http://www.readynas.com/wp-content/uploads/2008/09/ReadyNAS-Teaming.pdf. And they refer to "managed switches"...

[Leathal]

What is the best practise for best upload and download?



Ty

The DNS-345 doesn't properly bound, instead of using a virtual NIC and bonding both NICs together grabbing 1 IP address from the DHCP server both NICs grab an IP address and who knows what it does for bounding as it doesn't seem to do what industry standards saying bounding is.