Understanding throughput


#1

G’day setup a 1.1km B24 link… seems to be working great. Had to goto 2.5.2-ptp-beta5 FW because of assoc/disassoc issue, so far so good… the unit fluctuates between

PHY 1560 and 1300
and MAC 936 and 780

Have AUTO TDMA setup.

When I do a UDP bandwidth test across the link it can hit 900Mbps in either direction singly, but the web GUI of the B24 shows a MAC of 780Mbps…

Trying to reconcile and understand why I am able to exceed the MAC Mbps… Since this is Auto TDMA, is the actual throughput a combination of the Mac Tx/Rx (limited by Ethernet Port really)… if I had bi-directional traffic if both MAC Tx/Rx where at 780Mbps, then I could in theory pull down 900Mbps in one direction with 660Mbps in the other direction?)


#2

So, the B line of radios I have found to act a little funny about how they report their throughput ability. (best is to test with iPerf I have found, PITA to do a duplex test, but then you will know)

My B11s like to under report what they can do when there is not much bandwidth going through them. (A long conversation had just a little while ago about MCS stuff can be found here:General Mimosa rant) I have found that the B11s won’t go up to higher MCS rates unless they are under load. (This is probably a side effect of our “Conservative” setting for Rate Adaptation (found under “Link” in the UI) So you are probably just seeing the B24 report a lower rate then what it negotiates when the link is loaded.

Secondly. The B24 is not a full duplex device. So it will adjust how it times which side of the link is transmitting vs receiving according to which side has more to say. (Unless you hard set it in your link.) You are only going to see a maximum throughput of 1.7 Gbps PHY through the device, in very good conditions. For your link, as each side is lower then the max PHY you are going to see less. I can’t give a good guesstimate, but I would figure 1.1-1.4 Gbps, maybe. That one side is lower then the other will drag the whole link down. (because it will take longer for the slower side to send it’s information thereby increasing the amount of time it needs to send the info.)

I would do a run through your link and figure out where the problems might lie, in a PTP setup you should have almost exactly the same signal on each side unless there is something wrong. (More noise on one side of the link, missalignment, differing power levels, link obstructions and a whole laundry list of other things which have been mentioned in other posts.)