Dual Wan (Gold) - Not always load balancing for computer
I have two ISP's, 50meg down/10meg up. Load balancing on download I am seeing 80-90meg down (this is good).
However, it seems to come and go, and in some cases become locked at the max of one pipe (50down/10up)
I have found that by forcing a Queue Rule to limit to say 80megs/10megs (knowing I cant really LB on the upload)
It seems to refresh something, works for a bit...then backs off down to one pipe, in around the 40meg down
There are no other devices on it, about 3 computers as I am just tuning before full production mode.
Thoughts?
-
Firewalla load balancing is flow-based. So, if a larger flow may end up in one WAN and the next one will go to the next WAN. (the probability is based on the weight that you configured). so likely these flows some one lasting much longer and ended up (by chance) to one of the circuirts
-
That makes sense, however this flow is just a simple speed test, to speedtest.net
I have observed the speed test reporting 80+mb down, several times throughout a day....then the next day, down to less than one of the circuits. Both circuits are balanced 50/50 %
While testing with three machines, doing the test one at a time I see utilization in the 80meg download (that is good), then when doing two systems, same test, half that...adding a third I get a third...to be expected.
Now with one system and no other tests, just it alone using dual wan, it just doesn't gain the aggregated bandwidth.
Its no big deal, failover works, etc...but just strange that it seems to have locked on to this machine. Maybe a cache of sorts. (because after a reboot of the firewalla) it gives me full dual bandwidth, then on the 3rd to 4th test, it backs it down to the throughput on one pipe.
I just lack the visibility of what the traffic is doing, tcp flows, etc.
-
Agreed. Just sharing the behaviour mainly, as I do see positive results while with my personal laptop, all the time achieving excellent performance with combined wan throughput.
The other machine seems fixated. Both in the same VLAN, but a few times it achieved max throughput (combined)
What I observed, is when the machine was seen on two different vlan's, that by deleting the "offline" object, waiting a bit I would get positive results, so maybe some arp cache of sorts.
Anyway, the device is an amazing powerhouse and working extremely well.
Keep up the great work over there.
Please sign in to leave a comment.
Comments
4 comments