Has anyone had any success bonding 32 SC-QAM + 2 OFDM and achieving 2Gb+ download speeds? I am seeing poor download performance when testing this setup. If we bond just 2 of these channels in any combination, 2Gb+ speeds are achieved. Testing has been done on an Arris TG9452 and Hitron 5x1x modems with various firmware revisions with the same results. This has been tested on 2 different version of Harmonic CMTS (containerized and non-containerized) to Harmonic RPHY nodes in the lab with 2K and 4K modulations at the modems.
I have also configured our CBR-8 (RF linecards) for 2xOFDM channels but I am not sure if I am missing something. Our VeEx CX310 meter will connect on the downstream with 32+2 but our D3.1 modems are only bonding 32+1. The CBR-8 can only do 1xOFDM per DS port so I have tied a second controller in for OFDM#2 on the wideband interface. Software upgrades to the CBR-8 are not an option at this point in time and only D3.0 services are being provided to the remaining customers.
Any thoughts or suggestions would be greatly appreciated. Really just looking to prove it is the CMTS or the modem vendors and maybe it is both.
Thanks
I have tested it with Casa Core/Casa Nodes and Casa Core/Vecima nodes and it does work, but it is a bit tricky.
A few questions:
How are you testing it? Speedtest.net, iperf, other? What kind of speeds are you seeing when you have all three?
Does it work with UDP?
What is your upload set to? for TCP you will probably need at least a 150 Meg upload speed to acheive 2.5 Gbps for acks alone.
Most Core's I assume Harmonic as well, prioritize D3.1 interfaces for traffic meaning if you are bonded to two OFDM channels it will use both OFDM channels before using the 32 SC-QAM channels. What happens if you remove the 32 SC-QAM channels completely and just use the 2 OFDM channels?
Are there any FEC errors on the OFDM channels/SC-QAM channels?
Some Core's (read Cisco) require propietary Sid cluster settings to achieve high throughput, but they generally have the opposite effect on other Core's
Do you have the proprietary SID cluster settings on the CBR8? Do you have the same on the Harmonic?
I find it a lot easier to test all D3.1 first i.e. OFDM/OFDMA only get it working, then add the D3.0 channels in.
1. Testing via on-net Ookla servers. I haven't made it to iperf3 testing yet. Speeds vary between the CMTS'. CMTS#1 1700-2000Mbps/380Mbps and CMTS#2 800 - 1900/220Mbps. I have not tested with UDP.
2. Downstream/Upstream does not have a Max Sustain Rate set. Currently only max burst is set for both. I have also tried putting in 3Gb/400Mb in those fields. There is no change in the test results.
3. Harmonic balances traffic across all downstreams (SC-QAM and OFDM) until a default utilization threshold is hit on the SC-QAM, then it will start pushing D3.1 traffic to OFDM only. I am not sure what that threshold value is but I would guess around 60-65% SC-QAM utilization. If I bond any combination of the two then I can achieve 2Gbps speeds consistently. 32+OFDM1, 32+OFDM2 or OFDM1+OFDM2. The speeds start getting erratic only when all 3 are bonded together.
4. Based off of the cable meter we are running 1e -9 both pre and post for SC-QAM and OFDMs. I would assume there is potentialfor some correcteds when doing a speedtest at 4K QAM. OFDMA has shown 8% corrected at 2K QAM during a few test based on analytics.
5. Sid cluster is not configured on the CBR8 or Harmonic.
Update on the CBR-8 - I was able to achieve consistent 2.3-2.4Gbps download test on the Arris TG9452 with 32+2 downstream bonding. Nothing has been changed from the original post except finally getting a chance to connect the Arris modem. Current Arris config has no max sustain rates and does have a max burst set. The Hitron 5x1x will only connect at 32+1 and has no IP connectivity. Cannot ping CM IP from CMTS etc. I have never encountered an issue like this before. IP addresses are valid, no IP conflicts etc. The modem looks to be in working order from CMTS CLI. Could be the new config file that I am testing on the Hitron that the CBR8 doesn't like.
For this problem you definitely want a UDP testing mechanism
TCP adds a lot for extra unknowns that you can fight, i.e. is TCP window large enough for the latency at the speed I require
UDP will just work.
The other factor is the Web browser for ookla, what OS are you running it on? What browser?
The company I worked with used something like this: https://www.netally.com/products/linkrunner10g/ to do the testing as it had UDP and TCP options and did not require a browser
They had very hit/miss with Hitron Modems as well and had much better luck with the Arris modems and Sercomm modems
Have you eliminated the 2k/4k as the problem, 32 SC-QAM + 2x192 MHz at 1024qam should be enough for 2+ Gigs on the downstream.
The engagements I worked with needed the newest node firmware before 2k/4k worked properly.