1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

MM1000 Bonded MoCA?

Discussion in 'TiVo Help Center' started by tapokata, Jun 15, 2018.

  1. Jun 15, 2018 #1 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    I'm looking at upgrading the backbone of my coax plant. Current MoCA configuration is pretty basic- three drops- one for a Bolt, one for MiniVox, and the third as the ECB at the router. Using a TiVO ECB, the Bolt and MiniVox can run at MoCA 2.0 rates. My transmit power from all three nodes is -27dBm: the devices are not working that hard. PHY link rates are ~ 670 mbps (around 480 mbps MAC rate).

    I am in the process of adding three Motorola MM1000 Bonded 2.0 MoCA adapters, running ethernet from those to Bolt and the MiniVox. Using a different subnet configuration, you are able to log into the MM1000 adapter to pick up link rate details. With two of the adapters installed, I was curious what the new link rates would be, but I find no improvement in the PHY rate with only two of the adapters on the network.

    I thought, at first, that running the Bolt in MoCA mode was dropping the speed for the entire MoCA network (that shouldn't be the case), so I powered down the Bolt, and polled the rates with just the two new devices connected. The PHY rates and transmit power is essentially the same as the unbonded 2.0 devices (see attached screen capture).

    Am I off-base here to expect that the reported PHY rates will increase with the use of bonded devices? Or is the rate only showing for one channel, not combined?

    Any ideas? The devices are set for Band D extended, running on RF channel 1150. The Coax plant has a one 3x1 splitter, a Holland brand MoCA rated splitter. I can post a diagram of the plant, but I think most folks here have seen it.

    If I can't get bonded rate link speeds, I'm going to return the MM1K's- I'd just assume not have to open the third and install it, if the speed improvements aren't going to happen.

    As always, thanks in advance.
     

    Attached Files:

  2. Jun 15, 2018 #2 of 102
    krkaufman

    krkaufman TDL shepherd

    11,314
    1,423
    Nov 25, 2003
    The former is my impression ... the PHY rate value is just for a single channel; however, notice the bandwidth graphic to the right, which indicates two channels are in-use.

    To compare, drop-down to using just a single MM1000 in the setup and check the MM1000's stats page, again.

    p.s. Were you able to execute a bandwidth test, to determine the effective throughput (MAC rate)? I was a little surprised by the 480 Mbps rate you mentioned in standard MoCA 2.0 mode; I thought the max was about 400 Mbps, with extended/bonded MoCA 2.0 hitting 800 Mbps (putting aside Turbo mode's 1000 Mbps).
     
  3. Jun 15, 2018 #3 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    I think you are correct- it's only showing a single channel rate. The stats don't change dropping down to a single device.

    As for throughput, I was using a theoretical MAC rate yield, of about 72% of the PHY rate. But you got me thinking, and there's no time like retirement time for testing... so here goes.

    The bonded rates are faster... just how much, I'm not certain. I don't have another GbE ported computer in the house, other than my MacBook Air. What I do have in the plant, however, is a Netgear router, configured as a wireless access point (routing services are handled by an EdgeRouter Lite), and I'm using it's GbE LAN ports as a switch (as an aside, I am only using the LAN ports, not the WAN and LAN port, as Netgear has a bug/feature in their software where the LAN-to-WAN rates are slower than WAN-to-WAN). The ECB hangs off one of those ports. On that WAP, I have a shared USB 3.0 drive.

    I used Blackmagic, a freeware app primarily designed to test systems for video file transfer. The app creates a dummy file, which the user can vary in size from 1 Gb to 5 Gb. The app writes to and reads from a user specified target drive, every 8 seconds, and provides a couple of on-screen speedometers to show those rates.

    With the complete understanding that their may be overhead associated with the WAP switching, and that USB 3.0 drives, while rated at speeds up to 5 Gbps rarely hit more than 800 Mbps.. etc, I still have something of an apples to apples comparison for the adapter data speeds.

    As a benchmark, the results from running a GbE cat6 connection from the MacBook to the WAP LAN port:

    [​IMG]

    This is a capture from the app, cropped. 71.4 MB/s ~ 570 Mbps write speed for a 5 Gb file.

    With a MM-1000 as the bridge, and connected to the WAP switch, I then installed the TiVO ECB adapter (a MoCA 2.0 device) at another coax drop point, and tested again:

    [​IMG]

    That's a throughput speed of about 334 Mbps, on what is a MoCA 2.0 connection.

    Finally, I removed the TiVo ECB, and installed a second MM-1000 at that drop point, and ran the test again:

    [​IMG]

    That's about 492 Mbps for the MoCA 2.0 Bonded rate; or about 86% of the ethernet rate, but 47% faster than the MoCA 2.0 rate. I realize that's not laboratory precise, and there's probably some big flaws in this test, and the results are only comparable on my network configuration, but it's the best I could come up with while using the same fruit.

    As for why the link rates don't appear to show the bonded levels, I'm going to connect to MM-1000's directly to each other, using about 20' of RG6, and will check the reported PHY rates. If there's no change from what I'm seeing in the plant, I may ping Motorola support to see if they can provide an answer for the disparity.
     
  4. Jun 15, 2018 #4 of 102
    krkaufman

    krkaufman TDL shepherd

    11,314
    1,423
    Nov 25, 2003
    Yeah, a couple GigE-capable computers and a LAN speed testing tool on either side of the connection would be a more reliable test of just the network link. (See: LAN Speed Test, iPerf, jPerf)
     
  5. Jun 15, 2018 #5 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    True. Gotta go rattle around in the closet... I have an older HP laptop, is was upgraded to Windows 8, but I don't know if it's got a GbE port.

    FYI that I'm currently testing two MM-1000's on ~20' of RG6, straight cabled on the same bench. The reported PHY rate is 670 Mbps, with transmit speeds of -27 dBM. Those are the numbers I was seeing using the Tivo ECB.
     
  6. Jun 15, 2018 #6 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    Hmm. Port on win8 HP lappie appears to be GbE. Running LAN Test lite now.. results TK.
     
  7. Jun 15, 2018 #7 of 102
    krkaufman

    krkaufman TDL shepherd

    11,314
    1,423
    Nov 25, 2003
    As suggested before, the PHY rate will likely be the same as for the standard MoCA 2.0 adapter ... but two "bonded" channels means double the effective throughput that would have been possible with the TiVo Bridge. Also, I expect you meant "TX power estimate" of -27 dBm, rather than transmit speed; and -30 dBm is the max theoretical power reduction, with -27 the best I've seen real world.

    I *would* expect a slight boost in PHY rate with the two MM1000 adapters directly connected, since that should achieve the bonded MoCA 2.0 "Turbo" mode, capable of up to 1000 Mbps effective throughput.
     
  8. Jun 15, 2018 #8 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    yeah...TX power estimate. Brain fade from an old printer.

    I am in process of testing transfer times for 1MB, 10MB, 100MB, 500 MB, 1GB, and 5GB packet sizes, using LAN Speed LITE. Testing with a straight cable between computers as a baseline, then using MoCA with the two Mm-1000’s, and finally using the TiVo Bridge ECB in place of one of the MM-1000, which should be just a 2.0 connection.
     
  9. Jun 15, 2018 #9 of 102
    krkaufman

    krkaufman TDL shepherd

    11,314
    1,423
    Nov 25, 2003
    Keep in mind that LST Lite is subject to hard drive speed limitations, so it's not a true network speed test ... unless used in conjunction with the LST Server component.
     
  10. Jun 15, 2018 #10 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    Understood...

    The test beds were a MacBook Air running MacOS X 10.13.15, and an HP Pavilion DM-03 laptop, running Windows 8. LAN SPEED TEST Lite client was running on the MacBook. No internet connections were active, the only active network port was the GbE ethernet port on each machine. The TCP/IP connections were manually addressed, and no router services or controls were applied.

    No other processes were actively running, although given that Win8 may have a zillion different background processes running at any one time, basically the machines were in a real-world configuration. The HP is both old (about 10 years old) and slow (it was an XP machine in it's hey-day- and is clearly overmatched by Win8, although it does run). The MacBook Air, while not the youngest of chickens, has an SSD drive.

    I tested with a straight cat6 cable connection between clients, then added the MM1000 ECB's, connected via 20' of RG6 coax. The last test replaced one of the MM1000 ECB's with a Tivo Bridge adapter.

    [​IMG]
    As the packet sizes got smaller, the write times bounced around a bit- likely due to hardware, initial burst, etc. Over the larger packet sizes, the speeds seemed to stabilize, especially on the read times.

    The MM1000's, despite showing PHY link rates that are about the same as MoCA 2.0, are faster on data throughput- running about 82% of the direct cabled baseline, compared to 32% of the direct cabled rates in the standard baseline 2.0 configuration.

    For those keeping score at home, note that your mileage will vary, and these results are specific to my testing environment and equipment. While the conclusions can be applied in general, your specific results may not be comparable.

    I do have a support request in to Motorola on the PHY rate displayed in their diagnostics page, but in real world, the MM1000 appear to be running faster than a baseline 2.0 configuration, so I will let sleeping dogs rest for a while.

    I hadn't looked at the LST server mode- it took a bit of doing just to get the old HP to be able to share out a reachable folder on it's internal drive. Maybe for a later time, but my test window was about up, and I had to reinstall some components on the coax plant to keep the bride happy.
     
  11. Jun 16, 2018 #11 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    I registered and upgraded LST to the full version, and also purchased LST Server. I must be doing something wrong, because my Bonded MoCA read rates using LST server are about 20 Mbps slower than not using the server mode (the direct cabled rates improved, however, from the mid 230's to around 310 Mbps). I shut off all of the real-time virus and malware scanning on the LST client, too.

    Average read speed using the bonded MoCA adapters with LST Server was about 162 Mbps (testing each packet size 10x), or about 52% of the direct cabled rate.

    The developer states that a gigabit network connection should typically yield test throughput rates of around 600-800 Mbps with LST server, and 400-600 Mbps without it. I'm not getting close to that, even in a direct cable configuration. The 1.6 GHz Pentium laptop, with 4Gb of Ram running Win10, may be the anchor here, but I have no other device to test it on.
     
  12. Jun 17, 2018 #12 of 102
    Phantom Gremlin

    Phantom Gremlin Active Member

    1,575
    3
    Jun 20, 2002
    Tualatin,...
    I'm thinking of buying a few of these MM1000 adapters for my network, so this is an interesting thread.

    How is status info from the MM1000 accessed? The "User Guide" I downloaded from their website didn't mention any web interface or anything like that. How are you able to tell it to use a certain channel?
    Download User Guide


    Why do you need these bonded speeds? Your network as you have described it doesn't appear to need the higher speed.

    I plan on buying the MM1000's because I also want to push some non-TiVo stuff thru my MoCA network and don't want to rely on the built-in bridging in the TiVos to use the coax. But in my setup I'd be happy with MoCA 2.0 speed; bonded operation would be "nice" but wouldn't really be needed.


    I don't know that software, but is there any way to configure it to just do memory <--> memory transfers between machines? Once bits are being moved to/from hard disks or even SSDs, transfers probably slow down a lot.

    By the way, the Motorola of old doesn't exist anymore and I don't think it's the manufacturer of the MM1000.
    Zoom(R) Expands Motorola Brand License to Include Cellular and DSL Modems and Gateways, MoCA Adapters and Cellular Home Sensors

    So when you might "ping Motorola support" just know you're dealing with Zoom Telephonics. I don't know if that's better or worse than the old Motorola, but it's certainly different.
     
  13. Jun 17, 2018 #13 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    That information is found in a separate document on the Motorola support site, related to loading software updates. Connect directly to the adapter (and this can be done from a network connection, as well) change the TCP/IP IV4 LAN address of the client machine you are logging in with to 192.168.0.5, with a subnet of 255.255.255.0; and a router address of 192.168.0.2. Apply those changes to update the client LAN address, and from a web browser, enter 192.168.0.2, and you'll get a configuration menu for the adapter. If you have multiple adapters on the network, you'll be able to see all MoCA nodes on one screen, but you can only make changes to the last adapter that connected (so to apply a software update or make other changes, you need to connect directly to the ethernet port on the adapter). The configuration page has screens for entering security passwords (if desired) for the MoCA network, another page showing details for the adapter, such as firmware version, serial number, MAC address, etc. There is another page that shows the Node information on the MoCA network, including PHY connection rates from that adapter to others on the network. There is a statistics page and log file page, and there is another page that allows you to select the MoCA band and channel.

    Initially, I had the Bolt, Mini and TiVo bridge on a separate subnet, fed from my EdgeRouter. Their is no need of these devices to communicate with anything else on the network, so it's all good. However, I recently decided to take advantage of the unsupported bridging capability form the Bolt, adding a Netgear R6200 router as a wireless access point (running 5GHz AC only, on a different channel from the R7800 feeding 2.4 and 5GHz wifi to the rest of the house). I'm using the Gigabit LAN ports in the R6200 to connect to the MM1000, the Bolt, and an Amazon fireTV box for that room. Like you, I'm not confident in hanging all of that off of a bridged ethernet port, with a lower MoCA 2.0 data rate, so that was one of the reason's for the MM1000.

    As for bonded speeds? Because I can, I guess. The MM1000 is $20 less expensive than the Tivo Bridge 2.0 ECB. Where the Bolt is currently located, I don't have ethernet. I do have a fast ethernet connection to another room, where the Mini Vox is located, but it's 100 Mbps max (it uses only half the pairs of the cat5e cable, as the others are purposed for a previously but now unused IP based cable tv service), so MoCA to the MiniVox was the only way to get faster speeds than fast ethernet. Are those fast speeds necessary for TiVo? If I had multiple Minis attached to Bolt, then yes, but MoCA 2.0 is likely more than sufficient. The other devices in that location (tv, fireTv box, blu ray player) are all 10/100 ports, and don't access data concurrently, so the fast ethernet on a fast ethernet switch is sufficient.

    As KRKaufman noted, LST has a separate server package (LST Server), that for $6 for a single license can be loaded on any windows based machine- it does memory to memory transfers, so it is indeed faster. That said, the only other GbE port computer I own is that old HP Laptop with Win10- and the rates running LST-server on that machine are worse than writing to the shared USB drive... something to do with the age of the box, of the configuration. Direct cabled from a MacBook to the r7800 WAP where the USB drive is installed, a 300 MB test packet can be read from that shared drive at about 440 Mbps, as reported by LST on the MacBook client. Using the same cable and port and shared directory with the HP Win10 machine, and the read speeds on the HP are 150 Mbps. o_O. Something is goofy with that machine. I wish I had access to another windows based machine with a GbE port... but that's how it goes.

    Interesting. The packaging and documentation all refer to the brand and logo as being licensed to MTRL, LLC- a holding company with rights to the trademark. FWIW, I did get a response back on if I can get configuration information for a specific adapter in a network chain (answer- no), but haven't yet heard back on my query about the PHY rates shown on the Nodes page. We'll see. These adapters appear to be faster in throughput than the Tivo Bridge ECB, which is a MoCA 2.0 device.
     
  14. Jun 17, 2018 #14 of 102
    Phantom Gremlin

    Phantom Gremlin Active Member

    1,575
    3
    Jun 20, 2002
    Tualatin,...

    Thank you for that very detailed response.

    The MM1000 adapters are so inexpensive. I'm really tempted to pull the trigger and buy a few, even though things are working well right now. And I already have so much electronic "stuff" in the house. :)
     
    Mikeguy likes this.
  15. Jun 17, 2018 #15 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    I spent some time today attempting to clean up the HP win10 box- everything from applying numerous overdue updates overnight, to disabling a bunch of programs that loaded on start-up, likely hogging some of the 4 Gb of available RAM. After some work, I got LAN SPEEDTEST - SERVER ("LST-S" from here on out) running again. The client software, LST 4.0 ("LST") is running on a MacBook Air. I connected both computers directly to their respective GbE ports (the MacBook via a Thunderbolt connector) using a Cat6 cable. I ran the file transfer times for both write ("upload") and read ("download"), using packet sizes of 1 MB, 10 MB, 100 MB, 500 MB, 1 GB, and 5 GB. The packet size adjustment in LST allows for a fixed data length before acknowledgement receipt is required. Basically, the larger the packet size, the more constant the stream, although this is a TCP test, so there is error correction utilized. The chunk size of the data can also be adjusted, but I left that fixed at the default.

    Each test transfer was repeated 10x for each packet size, and I recorded the max rate for a given packet in the 10x series, along with the average rate for all tests for each packet size.

    The LST help files indicate a gigabit connection will typically show LST data transfer (using LST-S) of 600 to 800 Mbps. The results are comparable within the application, although the rates don't correspond with what you might find from iperf or other tests, but it's a good gauge of relative performance.

    I tested next running the server and client connected through the LAN ports of my Netgear R7800 router, which is configured as a Wireless Access Point. My EdgeRouter Lite is connected to another LAN port (not the WAN port) on the same router. The test results were within 1% of each other, so the baseline I elected to use was the LAN port configuration.

    I then connected the client to the Gigabit ethernet port on the back of one of the MM-1000 MoCA bonded 2.0 ECB's in the coax plant. Essentially, the test data from the client would be bridged to coax by the ECB, then bridged back to ethernet by another MM-1000, which was connected to a LAN port on the R7800.

    Finally, I removed the MM-1000 used as a client test point, and replaced it with a Tivo Bridge MoCA 2.0 adapter.

    So, the table of results:

    [​IMG]

    The good news was the the write (upload) speeds to the LST-Server box were over 800 Mbps, which was per the software developer is the expected "typical" throughput for a GbE connection. This was far better than my experience a few days ago, so that was a pleasant improvement. The MM-1000's provided maximum write speeds of 156 Mbps, and Read speeds of 181 Mbps. These speeds are 17% and 40% of the respective baseline. That said, the MM-1000 to Tivo Bridge rates, which would use an unbonded channel, is a little bit less than half of the data rates from the bonded 2.0 configuration. So in a relative sense, the MM-100's are operating at a faster data speed than a MoCA 2.0 configuration. (Sharp eyed observers will note that the Bridge configuration was tested out only to 500 MB packet sizes. I declined to test the the GB packet sizes, as read/writing 20 GB and 100 GB of data at the MoCA 2.0 Bridge rates is a thrill matched only by watching paint dry).

    One thing that I noticed is that the mm-1000's are using more power- the transmit power is about -16 dBM, compared to a low of -27 dBM from the TiVo Bridge. Note the statistics from the mixed device network:

    [​IMG]
    Node 0 is the MM-1000 adapter at the router; Node 2 is the TiVo 2.0 ECB adapter. Compare to the node info from the plan using all three MM-1000 adapters:

    [​IMG]

    When the Coax plant was at 2.0 speeds, using the TiVo Bridge, Bolt (as a MoCA client), and Tivo Mini Vox, TX speeds for all three nodes were at -27 dBM. The only devices changed in MoCA coax plant have been the MM-1000's. I'm curious if the TX power dBM rate reflects the bonded channel requirements in some way... which of course is further confused by the RX PHY rates NOT reflecting a link speed expected for a bonded 2.0 connection.

    As another curious aside, you'll note that the diagnostics web pages are nearly identical to what some have posted from their Actiontec devices: only the banners and logos are different- so I begin to suspect that Actiontec and Zoom-o-rola are using what is essentially the same innards for two different shaped devices.

    The TL/DR version: are the MM-1000's faster than the Bridge ECB? Yes, by a relative factor of about 2x. Do they even approach GbE speeds? Relatively, not.

    I suppose I'd like to have a proper Windows based server, one with a faster processor and oodles of RAM to test this out a bit more, but I'd like to have beer milkshakes for breakfast, too. Neither are likely. And so it goes.

    Are the MM-1000's a worthy investment? The answer is "probably." They are at a lower price point (currently) than a Actiontec or Tivo Branded 2.0 ECB adapter. Your mileage, as always, may vary.
     
    Last edited: Jun 17, 2018
    snerd likes this.
  16. Jun 18, 2018 #16 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    I was able to get iperf3 running on both machines (running as a server in the Windows box, client on the MacBook). I tested by running 10 consecutive sessions and recorded the results.

    Direct cable connection (Client -> r7800 LAN ports -> Server): 907 Mbps average, 912 Mbps maximum
    MoCA connection
    (Client -> MM-1000 -> coax -> MM-1000 -> r7800 LAN ports -> Server): 369 Mbs, 381 Mbps max.

    Given that iperf appears to run in memory, I don't think the hardware used as a server is a huge influence on the results- in any case, it's the same hardware, just connected differently. The throughput as a percentage of the direct cable connection is about 40%, or nearly the same as the read rates using LST-LST server.

    The MoCA results are what I would expect from a MoCA 2.0 connection :confused:?

    I'll need to remove the MM-1000's, and restore the coax plant (TiVo Bridge adapter, Bolt as MoCA client with bridged ethernet), and test the throughput by connecting the client at the Bolt's ethernet port.

    Project for a later time...
     
  17. Jun 18, 2018 #17 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    I thought I had an epiphany that the MM-1000's, which are default configured on a frequency of 1150 MHz, might require a higher frequency. Non-bonded 2.0 devices typically use the frequency range of 1150 - 1375 MHz, while bonded devices use 1400 - 1675 MHz. I disconnected two of the MM-100's and returned them to the bench- cabled them together with 20' of RG6 coax, and mounted the client and server PC's using manual TCIP addresses. I tried 1400, 1500, and 1600 MHz, using the D-High, D-Low, and D-extended options, but the iperf results are the same.

    When the bride isn't around, and I can get free access to the entire plant, I'll put the Bolt back on as MoCA client, and check the iperf data using the TiVo Bridge ECB, bridged to the Bolt's ethernet- as it was before I added the MM-1000's.
     
  18. Jun 18, 2018 #18 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    So, that moment presented itself. I removed the MM-1000, and restored the Bolt as a MoCA client. I bridged off the ethernet port, and running the iperf client:

    [ ID] Interval Transfer Bitrate

    [ 5] 0.00-1.00 sec 40.2 MBytes 337 Mbits/sec

    [ 5] 1.00-2.00 sec 42.1 MBytes 353 Mbits/sec

    [ 5] 2.00-3.00 sec 41.6 MBytes 349 Mbits/sec

    [ 5] 3.00-4.00 sec 40.2 MBytes 337 Mbits/sec

    [ 5] 4.00-5.00 sec 40.1 MBytes 336 Mbits/sec

    [ 5] 5.00-6.00 sec 39.6 MBytes 332 Mbits/sec

    [ 5] 6.00-7.00 sec 40.1 MBytes 336 Mbits/sec

    [ 5] 7.00-8.00 sec 39.8 MBytes 334 Mbits/sec

    [ 5] 8.00-9.00 sec 39.9 MBytes 335 Mbits/sec

    [ 5] 9.00-10.00 sec 39.7 MBytes 333 Mbits/sec

    - - - - - - - - - - - - - - - - - - - - - - - - -

    [ ID] Interval Transfer Bitrate

    [ 5] 0.00-10.00 sec 403 MBytes 338 Mbits/sec sender

    [ 5] 0.00-10.00 sec 403 MBytes 338 Mbits/sec receiver

    These are very similar to the iperf reported rates running two of the MM-1000's directly connected together, with no splitters, switches, routers, etc, and no other MoCA devices on the network. I can only conclude that these adapters can't deliver MoCA 2.0 bonded speeds.

    They'll be going back to Amazon.

    Thanks for listening!
     
  19. Jun 18, 2018 #19 of 102
    Phantom Gremlin

    Phantom Gremlin Active Member

    1,575
    3
    Jun 20, 2002
    Tualatin,...
    From the numbers you give, I don't think your adapters are doing the "bonding" thing, they're just running plain MoCA 2.0. Bonding takes a total of 225 MHz. 100 MHz + 25 MHz gap + 100 MHz. Here's a document that shows that:
    www.mocalliance.org/MoCA2/specification/MoCA_2.0_and_2.5_Device_RF_Characteristics-160808d.pdf

    Of course they should be bonding, the MM1000 blurb says
    • Speeds up to 1,000 Mbps!

    The exclamation point is theirs, not mine.

    I would disconnect the adapters from both Ethernet and coax, but keep them powered up. Then reset the adapters (use the hardware button on each). Then power cycle them again. Then hook them up directly with the 20' of coax. Don't try to configure them, don't go into that menu. Let them negotiate on their own, starting with a factory fresh configuration. It seems weird that they can't figure out to do bonding properly.

    As to why their GUIs seem similar, both manufacturers are probably using software provided by the SOC manufacturer, which IIRC is Broadcom.

    Edit: do you have the security feature enabled? The one that requires a 17 digit numeric code? Maybe the adapters can't run at full speed if they're doing encryption?
     
  20. Jun 18, 2018 #20 of 102
    tapokata

    tapokata Active Member

    241
    74
    Apr 26, 2017
    Sacramento, CA
    Thanks, but I did all that you suggest (full reset, power cycle, hook up to straight coax) after realizing that the frequency changes weren't making a difference (and to make sure that I hadn't bolloxed something up making frequency changes). Even after a factory reset, power fail and then allowing things to settle down for a few minutes before testing- there was very little change in the iperf rates from what I posted in reply #16, and those posted rates reflected the connection through the LAN ports of the netgear R7800 wireless access point.

    This final testing of the MM-1000's was done on the bench- 20' of RG6 coax connecting the adapters; the supplied ethernet cables cabled directly to the GbE ports of the client and server machines. IPV/4 addresses were entered manually (198.168.0.5 for the client, 198.168.0.7 for the server, subnet of 255.255.255.0 for both). I used the latest versions of iperf for MacOSX (the client) and Windows10 (the server). I even used the -R qualifier in the command line from time to time, to send from the server back to the client, looking for a difference (there wasn't one). Again, the same client and server boxes on an ethernet cable, without the MoCA adapters, yielded rates in the low 900's, so the client and server hardware aren't the issue.

    Edit Response: No security key was entered or used on any adapter.

    I was as gobsmacked as anyone else when testing my original configuration, the MAC rates from a non-bonded 2.0 network were about the same as the "bonded" units.

    If these units were bonding, they weren't showing it in the throughput. Perhaps they've just grown apart, and aren't the soulmates they once were. I dunno. The TX power rates were -16 dBM, as reported by the software- the RX PHY link rates were 642 Mbps, as reported by the software (again, speculation by others here is that reported PHY rate is for the primary channel, not the combined). I don't know if I had one defective MM-1000 out of three; I guess I could have played round-robin with the three units to see if all three, attached two ways, provided the same result. I could have experimented more to see if two out of the three would find that etheral spark that paired them as universal constants on this mortal coil, allowing two of them, as many, to become more than the sum of their parts- but, alas, I didn't have it in me to identify and tear the heart out of the loser. The fault, you see, is not in our stars, but in our command shells. It's better that I didn't know.

    The bottom line is the tested MAC rates for these ECB adapters are practically the same as what I have going through my existing MoCA 2.0 Coax plant, using the TiVo Bridge (ECB 6000) and the Bolt as a MoCA client, connecting the iperf Client to the Bolt's GbE port.

    I wrapped each in a participation ribbon, boxed them up, and shipped them back a few hours ago. They are certainly a less expensive option to the TiVo Bridge as an ECB adapter choice, but I don't need a near $200 investment to replicate the speeds I already had with the TiVo MoCA 2.0 clients and the ECB.

    It's not a complete loss. I got to spend $16 on a couple of pieces of testing software (LANSpeed Test), as well as the joys of learning how to configure and run iperf (for free) on both a Mac and Windows machine. I also discovered that my dust-gathering vintage 2010 HP laptop had a GbE port, too. Who knew! As for the future of my network, maybe one day Actiontec will release the ECB6500 to the common man- with PHY rates of 2.5 Gbps; a virtual steam-shovel compared to the John Henry that is unbound MoCA 2.0. To sleep! Per chance, to dream!

    And so it goes.
     
    Last edited: Jun 18, 2018

Share This Page