Actiontec MI424WR - a cheap MoCA bridge for all! - Page 13

I have to look at the network monitor next time I stream the movies, both MPEG2 and H.264.. I used FTP and JPerf to test my file transfer speeds and they both run concurrent..


I have to get a HD video file though to stream..

Glad you got yourself straightened out, but do test that 2nd router soon.. I got 3 duds out of 6 from them, though they were very pleasant dealing with getting the exchanges set up..

As per ActionTec's response:

CM Ratio : Its a setting pre-determined by Verizon for connection to the Optical Network Terminal or ONT for the FIOS service. Its not used for any other types of connection. This is the modem or ONT connection for the WAN port of your router and the setting is programmed so that the ONT can provide the Internet data\\video connection on your FIOS services.

I don't have a WDTV device, but just a few google searches comes up with lots of people having problems with WDTV at HDTV resolutions when using windows file shares. Pretty much everywhere it seems, WDTV people recommend using NFS instead. I think you should give that a try.

I know its frustrating that it seems to work in a tight LAN situation, and doesn't work across the house, but its likely that windows file sharing just doesn't quite handle nicely the extra latency or slight differences that the actiontec is introducing.

It's definitely true that the WDTV doesn't handle streaming over the network as well as say WMP on the PC, but the behavior I saw seemed to be independent of the player used. There was a distinct bandwidth "shelf" for streaming (20mbs) which doesn't jive with the 85-90mbs I'm seeing transferring large files.

One thing that we have found is that the backplane of the actiontec routers might be a culprit. Even when both devices are connected directly to the switch I get stuttering. So my assumption is that the it has to be something within the actiontec itself that is not handling the streaming well outside of the conversion from coax to ethernet.

You asked for our help. Do the NFS share. Believe.

You got it my man. I'm downloading windows services for unix as I type this. For anyone that's interested, the download is here.

I'll report back what I find.

This isn't a guaranteed fix. I just wanted to see the results from the streaming using a different protocol. NFS is a pain to setup on Windows.

Your not kidding..

ok, just an update. i was not able to use windows services for unix since it is not supported on 64 bit architecture. I was able to find allegro which is free for the first 30 days. The problem is that the WDTV does not natively support nfs file shares. Looks like I will have to install a modded/hacked firmware to enable this feature. I have not had a chance to do that yet, although I've downloaded the B-RAD firmware. Once I get a chance, I'll report back. I'm starting to doubt if it was worth all the trouble...

I wouldn't bother with installing new firmware for this experiment. I thought you would have already been running 3rd party firmware on WDTV. At some point it just gets easier to run CAT5e...

Those settings for CM Ratio and Routing didn't work for me either. I'm beginning to suspect it's my coax cabling setup..

I run Windows services for UNIX (on XP) and using the NFS share, I'm not seeing any difference in my speeds (~8-15Mbit/s) on 720p x264 mkv's.. still get stuttering on some movies too (i.e. Toy Story 1 and 2).

It just might be your coax lines.. I would definitely try re-terminating those ends and maybe using different splitters.. Try to get the signal to pass as little amount of coaxial as possible.. If you get daisy chain them using a short run of coaxial for testing and get the same speeds, your Actiontec may be bad..

I was not able to get a 1080p MKV rip to play without stuttering over MoCA on a 90 Mb/s connection, so I believe the ActionTec has a flaw in its hardware design (Revisons A,C,D)..

Well, I tinkered with my coax a little bit with little success. I traced the coax to a drop amp and then removed that amp so the main run goes to a three-way splitter to another 3-way and then to my Actiontecs. No difference in speed.

As I mentioned earlier, I tried daisy-chaining two Actiontecs together with a short 3ft coax and from one laptop to another laptop via the Ethernet ports there was no difference in speed (still 10-20Mb/s) so now I'm back to thinking it's the routers which are having problems - something either with the Ethernet ports/built-in switch or the Ethernet to Coax conversion..

What I don't quite understand though is when I set the TCP window size to 100 kb, and re-run the iperf/jperf tests, I get ~60-70Mb/s through the coax, so that should tell me it's working (although not at the 90Mb/s you are getting) somewhat, correct?

Puzzling.. I'm going to try to do a full reset (via the web interface) on two Actiontecs and possibly update the firmware on my Rev A's to see if that helps any..

Can any of you with Rev A's please report your firmware version?

Mine is currently 4.0.16.1.45.160

Thanks,
Chad

vegitacx had the same problems with both his Actiontecs, bought two other routers and the problem was solved.

If you do go the firmware route, the latest firmware is 4.0.16.1.56.0.10.11.6..

I had the same problem with one of my routers being "stuck" at 10mbit. The other router worked once I hard reset it a few times (just luck I guess). I bought 2 from ABCliquidators and both worked fine at 100mbit. My home network is all sailing now.

How do you hard reset it? Hold down the reset button for 10 secs on the back? Do you have the ports set to Auto or 100/Full?

Maybe it's time to buy some additional units..

Power on and after it boots hold reset button for 30secs, power off while continuing to hold reset button for 30secs, and power on while continuing holding reset button for 30secs.

Thanks, but this doesn't work for me. Once I have turned the power back on after the last 30 seconds, the power LED goes orange, and I can't do anything. I ended up just holding down the reset button for 10 seconds, waiting for all the lights to go out, and then re-configuring the box.

I went back to square one and daisy-chained two Actiontecs beside each other via coax.. still no change in speed (maxes out at ~15 mbps).

However, when I changed the TCP Window Size to 100 Kb and re-ran the jperf test, the speed jumped to ~60-80 mbps.. so I'm confused as to whether there is a problem here or not..?

(BTW, I did upgrade the firmware on all Rev A Actiontecs to 4.0.16.1.56.0.10.11.6-MI424WR.rmt)

I have no experience with jperf but I basically perform copies and ftp transfers from both ends with laptops to test the throughput. My second router that was maxing at 30mbit on streams, maxed out at 90mbit on local traffic so when I reset it something changed and I was getting 90mbit on streams.

It may be only one of the routers that's really limiting the whole thing. When I did a local traffic test on both routers one popped up with 90mbit (the one that fixed itself with hard reset) and the other hit 10mbit locally. No matter what I try I can't get more than 1.1meg/sec from that router.

If that's the case I would just buy a couple from that place, it's relatively cheap for the added functionality.

You probably have 1 or more bad routers.. If you cant return the ones you have, make sure the seller allows you to return them when you order the replacements.. When they work, they work very well..

Seems like I'm not the only one with the question about why adjusting TCP Window Size increases MoCA performance:

http://forums.smallnetbuilder.com/sh...2&postcount=83

Can anyone here explain this apart from return and replace?

Also found this on the amber light when trying to do a hard reset:

http://www.dslreports.com/forum/r198...WR-via-updates

..seems as though you have to use TFTP once the amber light appears, and then manually put the firmware file on it.

Using TFTP to install the firmware is only used should you brick your router and you cannot connect to the UI..

I did not adjust the window size settings with my routers and they operate at 100Mb/s speeds, so it seems your cheating if your using those settings to get speeds you cannot achieve without using them..

I have a 6ms ping time between two different hosts across MoCA.

With 6 ms, you need at least the following TCP Windows Size to get 100 Mbit

100000 kbps x 0.006 seconds / 8 bits/byte = 75 kbytes

The iperf server tcp windows size defaults to 8 kbytes on my windows computer. If you don't set it above 75 kbytes, you won't be able to push up to 100 Mbits/sec.

iperf default settings don't reflect what your default OS kernel or applications like ssh or nfs would be configured at.

This is a pretty broad question that I don't have the explicit expertise to answer authoritatively. Each OS has knobs that can and can't be turned and applications vary dramatically as to their TCP window size defaults and TCP window auto-scaling capabilities.

folks - sorry to interrupt the performance discussions... one of my two deployed boxes seems to have bitten the dust. as far as i can tell it's a hardware failure of the 4-port switch. when i plug any device into any of the 4 ports, the activity light flashes very quickly, and the interface drops almost every 64-byte packet sent to it. larger packets are dropped entirely.

has anyone else seen this? i did buy three of these to cover myself in this event, and i'm back up and running with the spare. but given that 1 out of the 3 that i bought initally was bad, and now that another has failed, i'm a little worried that maybe these boxes are not so reliable.

does anyone know how to open one of these? i wanted to take a look at the board and see if anything looks obviously blown...

edit - i got it open... 4 screws under the footpads and one under the sticker at the front. can't see anything obviously wrong. the power supply shows 5.2V unloaded and 4.9V loaded. that's probably okay, i suppose. i have heard that the power supplies do fail... i wonder if a bad power supply could have killed mine.

at any rate i need to seek out another spare...