SmallNetBuilder

Performance - Bench Throughput

I first set up the pair of MCAs on my bench, with one of the 1m RG6 coax cables supplied in the kit connected between the Coax In jacks, one of my trusty test desktops connected to one of the MCA's Ethernet jacks and the other MCA's Ethernet connected to another test machine via one of my LAN's switches.

I then ran a series of IxChariot tests through the pair using the throughput.scr, TCP/IP and changing only the test file size from its default of 100000 to 3000000 Bytes. I also ran reference tests using a 10/100 Ethernet connection between the two test machines.

Figure 9 shows a composite of transmit tests through the MoCA and Ethernet connections, where we see that only around 70 Mbps of the touted 270 Mbps is actually delivered as usable throughput. Receive results are virtually identical, but you can see for yourself here.

Figure 9: Comparison of MoCA and Ethernet transmit throughput

I next checked to see how a simultaneous transmit / receive test fared, which is shown in Figure 10, once again with Ethernet and MoCA results included. Both the Ethernet and MoCA results sacrifice some throughput in the individual receive and transmit connections. But note that aggregate MoCA throughput moved up to 110 Mbps.

Figure 10: Comparison of MoCA and Ethernet transmit / rcv throughput

I was curious to see if there was any more throughput to be gained from adding more connections to the test. So I ran test with up to 10 connections (5 transmit / receive pairs). Figure 11 shows the three connection test, with 128 Mbps of aggregate throughput.

Figure 11: Three connection throughput

And Figure 12 shows 4 connections, which provided just shy of 138 Mbps of aggregate throughput, which turns out to be the most that I could coax from the connection. Note, however, that Pair 2 has much higher variation and about 15 Mbps lower average throughput than the other three pair.

Figure 12: Four connection throughput

I poked at this a bit more and Figure 13 shows a 10 connection test. Although aggregate throughput remains up near 140 Mbps, clearly there is a battle for bandwidth occurring. The nice steady throughput found with 3 connections has turned into a mess more like what I see with draft 802.11n connections. I think you would be hard-pressed to get a reliable 720p HD stream with variation like that.

Figure 13: Ten connection throughput

I also ran some ping tests to check for latency. Figure 14 shows a consistent 3 to 4 ms ping time, which says latency won't be a problem.

Figure 14: MoCA direct connection ping test