Tomorrow morning I head out for a three-week stretch of near-continuous travel, so as a 'teaser' for Maury's broadband over powerline article which'll be published in a bit over a week, I fired up the DS2 powerline adapters (which I first mentioned a few weeks ago) yesterday and did some performance testing. As I previously suggested, before continuing I encourage you to read (or re-read) my short article 'A Man, A LAN, a Plan' originally published last August. Specifically, peruse the floorplan of my network's topology (link is to a PDF); it'll help make sense of what's to come.

To measure the transfer rate performance of the networking link created by the two DSS9001-based Ethernet-to-powerline adapters, I copied files between my Gigabit Ethernet-equipped Dell desktop PC and Infrant's ReadyNAS. A few qualifiers before continuing; first recall that as I've mentioned before, while my Gigabit Ethernet switch and the ReadyNAS both support jumbo frames, none of my Gigabit Ethernet-equipped computers offers this potentially performance-boosting feature. Also, I'm testing a two-adapter link, versus the more elaborate three-adapter powerline networking configuration I initially pursued, and over a comparatively robust segment of my home's powerline topology. I'm also doing my testing outside of the summertime period that, from past experience, particularly provokes powerline networking problems to manifest.

With the fine print out of the way, let the testing begin. I obtained my baseline set of datapoints after connecting the desktop PC to the ReadyNAS over CAT5+ cable, via my SMC Gigabit Ethernet switch. I'd obtained similar data during my initial testing of the ReadyNAS, but Infrant had swapped me units to resolve the front panel power switch problem I uncovered with the original system, and the company had also recently released a firmware upgrade, so I decided a re-test was in order. Writing my 6.8 GByte 'torture test' file, the ReadyNAS achieved around 90 Mbps peak speeds; while reading that same file back, the ReadyNAS pegged the meter at ~190 Mbps until the HDD buffers were depleted. Part of the reason why I delayed the testing of the DS2 adapters until now is because I was waiting for replacement drives from Maxtor and Seagate (which still haven't arrived) that include 16 MByte buffers. As you'll soon see, though, the DS2 adapters wouldn't have been able to exploit any added performance that the replacement ReadyNAS drives might have delivered.

Next step; some real-life testing of HomePlug v1.0 Turbo units, which tout up-to-85 Mbps PHY rates. I moved the ReadyNAS to the hut, direct-connected it to one of the Aztech adapters there, and connected the other HomePlug 1.0 Turbo adapter to my Gigabit Ethernet switch in the office. Essentially, I extended my previous tests' connection between the desktop PC and ReadyNAS via a powerline segment. Intellon's PowerPacket diagnostics utility reported a high quality, >60 Mbps link between the two adapters. However, when writing the 'torture test' file to the ReadyNAS, I could only achieve ~20 Mbps sustained speeds, and reads were even slower, at approx. 15 Mbps. This discrepancy reflects the earlier-mentioned asymmetrical bandwidth characteristic of my powerline network, as reported by the Cogency Connection Manager diagnostic utility I'd earlier used with 14 Mbps HomePlug 1.0 adapters.

Continued with 'Powerline Followup: Preliminary Results and Next Steps'