Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
cancel

ASK THE EXPERT- BROADBAND CABLE

Welcome to the Cisco Networking Professionals Ask the Expert conversation. This is an opportunity to discuss Broadband Cable with Cisco expert Falk Ludviksen. Falk’s current responsibilities include Technical Assistance Center (TAC) Escalation in which he troubleshoots complex issues, provides training, and authors documentation. He is also recognized as a Cable Worldwide Subject Matter Expert and often reviews and analyzes feedback to determine information needed by customers to solve issues. Feel free to post any questions relating to Broadband Cable. Remember to use the rating system to let Falk know if you’ve received an adequate response.

Falk might not be able to answer each question due to the volume expected during this event. Our moderators will post many of the unanswered questions in other discussion forums shortly after the event. This event lasts through February 14 Visit this forum often to view responses to your questions and the questions of other community members.

25 REPLIES
Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

I have to make a performance tests on the upstream path in a HFC network. The purpose of my test is to individuate how many fiber node I can combining together in one upstream port of CMTS.

CMTS: uBR Cisco 7246VXR (at the moment there is DOCSIS 1.0)

CM: Cisco uBR924

QPSK modulation

For the moment I use a software test (Qcheck) by NetIQ. This software is running both in my notebook connected to the cable modem and both in server connected to the CMTS. In detail my notebook is connected, through the HFC network, to one upstream port of CMTS where I have, through its Ethernet port, another PC server running Cisco network Registrar, remote administrator program and the software Qcheck.

I measured the throughput of TCP packet changing the size of the packets (100Kbyte, 500kbyte, 1000kbyte) and the RF bandwidth of upstream ( 200kHz, 400kHz, 800kHz, 1,6 MHz, 3,2 MHz ).

The result, considering 20 samples for each experiment, is that for the packet of 100Kbyte I can notice a variation of throughput in the 20 sample and that variation become bigger with the increase of the bandwidth. ( From 200kHz to 3,2 MHz). If I do the same test with packet of 1000Kbyte the results of 20 samples is that the range of their variation is less and when I increase the bandwidth each sample is very very close to the others.

About that test can I say that the small packet are more variable because of the Noise? If the packet is too short I have more probability that the noise can damage the packet? What kind of Noise I can see in the upstream path?

Another result is that with 200Khz I have the average bit rate of 238 Kbps, with 400kHz the average is 428 kbps, with 800 kHz the average is 732kbps, with 1,6 MHz the average is 1178Kbps and with 3,2 MHz the average is1555kbps.

In this analysis the ratio between the maximum raw bit rate (for each RF bandwidth) and the average bit rate decreases. Is this also related to the Noise ? and how?

The next step should be to repeat the same tests with more fiber node combined together and analyze how change the performance. For the moment I don’t have any kind of hardware device to measure, do you think that this procedure is right. I would be very pleased if you could give any kind of advices.

Thank You

Francesca

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

Francesca,

Hi! Regarding determining how many fiber nodes to combine at the input to a given CMTS upstream port, I think I'd approach it from a different perspective.

The DOCSIS Radio Frequency Interface Specification states that upstream carrier-to-noise, carrier-to-ingress power and carrier-to-interference ratios are supposed to be "not less than 25 dB." The first factor used to define the number of fiber nodes per upstream port should be the 25 dB "carrier-to-junk" parameter. I recommend that you measure each node's upstream C/N ratio, then do some simple power addition to determine the maximum number of nodes that can be combined without resulting in the combined C/N ratio dropping below the DOCSIS minimum of 25 dB. For added reliability, I'd set a goal of 30 dB C/N at each upstream port rather than the DOCSIS 25 dB number.

Let's say each node has 50 amplifiers, each with +18 dBmV input and 10 dB noise figure. The standalone C/N ratio for each upstream amplifier will be ~67 dB. The combined C/N ratio of 50 amplifiers, each with 67 dB C/N ratio, will be ~50 dB. Typical standalone upstream fiber link C/N is usually in the high 30s--let's use 39 dB for this example. When 39 dB is combined with 50 dB, the resulting C/N ratio is 38.7 dB.

The next question is how many nodes with this C/N ratio (38.7 dB) can be combined before the total C/N ratio drops to 30 dB? Two nodes, each at 38.7 dB, combined would be 35.7 dB. Four, each at 38.7 dB, would be 32.7 dB. Eight nodes, each at 38.7 dB, would combine to result in 29.7 dB C/N ratio. Backing off to seven gives us 30.25 dB combined C/N ratio.

The calculated C/N ratios in this example use thermal noise from each amplifier and the upstream optical link, and do NOT take into account the real world of ingress and impulse noise, so the real "carrier-to-junk" ratio likely would be quite a bit less than the numbers used here. I've yet to see a real-world CATV network that has only thermal noise. That's why I recommend the actual C/N ratio be measured.

The other factor that must be taken into account is the desired throughput versus number of cable modems per upstream port. Cisco's recommendation has been to keep the number of cable modems per upstream port somewhere in the 150 to 250 range.

When these two factors--network C/N ratio and number of cable modems per port--are taken into consideration, then you can determine a reasonable number of fiber nodes per upstream port. The number of fiber nodes is not the real issue here. What's far more important is the number of modems per upstream port in conjunction with the combined C/N performance. For reliable performance, the combined "carrier-to-junk" ratio MUST be better than the DOCSIS 25 dB spec, and the maximum number of cable modems per upstream port should be kept in the 150 to 250 range.

Sincerely, Falk

Falk Ludviksen

Cisco Systems Inc.

Cable & Voice TAC

Phone: 408-525-4379

AIM: fludviks

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

Thank You for your answer.

I realized that the procedure to analyse how many fiber node combine together in a upstream port was not right; I have to know the SNR of cable plant’s components. Even I would like to test data throughput in the upstream channel ( please correct me if I am wrong) it has non sense measure the throughput sending TCP or UDP packet with 1 Kbytes to 1000 Kbytes size. The problem is that I have only a software tests that could analyse data throughput through TCP and UDP packets. I need some measures for my thesis degree, I have the possibility to tests a real HFC network but I can use only that software tester. Could you advice me what kind of test I can make with this tester or do you know where I can find any result of reference performance tests using Cisco CMTS uBR 7246 VXR and Cisco cable modem?

This kind of test are very important for me, I WOULD BE VERY PLEASED if you could help me.

Best regards

Thank You

Francesca

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

Falk,

I understand the C/N approach to determining the number of nodes-per-upstream point. However, isn't there a customer limit per US port that one would want to consider as well? Assuming so, it would seem one would need to know how many homes were passed per node and the actual/projected penetration of CMTS modems. (I was taught not to put more than approx. 200 modems on an US port).

Also, what about other US services running in the 5-40 MHz spectrum. I've always wondered if there was an "total RF energy" limitation on a given US port (kind of like such that would clip a laser). Or would the reverse laser be the "weak link" in that respect?

Thanks,

Jeff Mills

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

I'm trying to configure a uBR7114 with the right balance of subscriber privacy/security and functionality. Current IOS is; (UBR7100-IK1S-M), Version 12.1(10)EC.

At present, ip proxy-arp is disabled creating the condition where 1 CM subscriber is not able to communicate with another CM subscriber on the same system. Enabling proxy-arp opens up the system to much in my opinion but in keeping things status-quo some subscribers who telecommute between home and office in the same town (both locations using cable internet) can't connect between each other. I've read that two subscribers may be able to manually add ARP entries for systems they wish to connect to, but when businesses/users have firewall routers, they aren't able to make such entries if in fact that works. Is there a better way to enable proxy-arp but yet filter out netbois (Windows File & Printer Sharing)? Is there a different solution altogether.

Thanks in advance,

David Garland

netINS

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

David,

Turn on "cable proxy-arp" on your uBR7114 so the cable modems can talk to each other. Then filter out netbois. There are two types of filtering , the Logical Link Control (LLC), and IP traffic. Filtering can be set in two ways:

1. DOCSIS Configuration File

2. SNMP

DOCSIS Config file tool sets IP and Logical Link Control (LLC) filters for your cable modems. For a detailed explanation on the OID's that you need and how to build the DOCSIS configuration file to filter netbois, please read the "IP and LLC Filters" sub section of How to build DOCSIS Configuration Files For DOCSIS 1.0 using Cisco DOCSIS CPE Configurator V3.2

http://www.cisco.com/warp/customer/209/docsis_config4.html#511694945

The SNMP method, you need to set the two following tables from the DOCS-CABLE-DEVICE-MIB:

docsDevFilerLLCTable for the LLC filtering

docsDevFilterIPTable for the IP traffic filtering

Sincerely, Falk

Falk Ludviksen

Cisco Systems Inc.

Cable & Voice TAC

Phone: 408-525-4379

AIM: fludviks

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

Thank you for the reply....

Are the IP and LLC filters that would be created through the DOCSIS Config file applicable to only Cisco Cable Modems or to all CMs?

Thank you,

David Garland

netINS

515-830-0534

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

David,

Hi. Filters apply to DOCSIS compliant cable modems.

Sincerely, Falk

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

Thank You for your answer.

I realized that the procedure to analyse how many fiber node combine together in a upstream port was not right; I have to know the SNR of cable plant’s components. Even I would like to test data throughput in the upstream channel ( please correct me if I am wrong) it has non sense measure the throughput sending TCP or UDP packet with 1 Kbytes to 1000 Kbytes size. The problem is that I have only a software tests that could analyse data throughput through TCP and UDP packets. I need some measures for my thesis degree, I have the possibility to tests a real HFC network but I can use only that software tester. Could you advice me what kind of test I can make with this tester or do you know where I can find any result of reference performance tests using Cisco CMTS uBR 7246 VXR and Cisco cable modem?

This kind of test are very important for me, I WOULD BE VERY PLEASED if you could help me.

Best regards

Thank You

Francesca

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

Francesca,

I suggest testing performance as described in this url:

http://www.cisco.com/warp/customer/109/troubleshooting_slow_perf.shtml

Also, to assist you on your Thesis, please read this white paper which I co- authored: "Understanding Data Throughput in a DOCSIS World" found at:

http://www.cisco.com/warp/customer/109/data_thruput_docsis_world_19220.shtml

These two urls should be extremely helpful.

Sincerely, Falk

Falk Ludviksen

Cisco Systems Inc.

Cable & Voice TAC

Phone: 408-525-4379

AIM: fludviks

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

Hi,

I have problem with choose software from 12.2 line on uBR7246 (CMTS 1.1 - CM 1.0). On CCO are available two IOS line BC and CX. But both have bug. Software BC doesn't support second queue (base on precedence) and CX have problem with showing errors on interface (hop and inf errors).

And one more thinks - the IOS must support NPE-G1. What do you suggest?

Which line is better use for 1.1 transmission? Which soft will be more development by Cisco?

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

Hi. The Cisco UBR7200-NPE-G1 is not supported on the Cisco uBR7246. You have to have a uBR7200 VXR chassis. In general however, I suggest the BC code as the main image over CX which has been through much more tests and deployed more frequently in the field. Both have DOCSIS 1.1 support. I see the focus on BC for the future. However, make sure you have a CX boot image. Sincerely, Falk

Falk Ludviksen

Cisco Systems Inc.

Cable & Voice TAC

Phone: 408-525-4379

AIM: fludviks

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

Hi. Yes, I have 7246VXR, but as I understood both image support NPE-G1. But indepedently forn CPU you suggest to use BC train? Do you know when will be second queue supported in BC? Someone work now on this bug?, because I don'y see it on Bug tool.

Community Member

Re: ASK THE EXPERT- BROADBAND CABLE

Hi. I'm unfamiliar of a feature called "second queue" and I do not see it in the release notes for BC or CX train. Is there another name for this feature? Can you describe it to me? Also, can you provide me the DDTS your refering to so I can better assist. Thank you very much. Sincerely, Falk

73
Views
5
Helpful
25
Replies
CreatePlease to create content