With unprecedented demand facing local broadband networks, operators are under pressure to prepare for congestion and capacity issues before they happen. In this blog I’ll share some tips that will help alleviate and mitigate congestion, while addressing capacity concerns on cable operator DOCSIS plants.
A Nielsen.com study on streaming video minutes on networks compared usage this year versus the same time last year. The study showed that usage increased 90%, from 69.8 million minutes per week to 161.4 million minutes per week. Now that’s a lot of binge-watching!
So, what are we all watching? The big increase is fitness programming and videos – up 147%. Life and home programming were also up with a traffic increase of 64%, and premium channel content was up 54%, while music was down 10% from this time last year.
How does your broadband connection handle all this new traffic?
To help cable operators keep up with the traffic demands, our operations and engineering teams are working ’round the clock taking fast actions. What can cable operators do now to tweak their cable modem termination systems (CMTSs) for the current traffic surge?
Seven Tips for Managing Capacity and Congestion
- Decrease subscribers per service group (SG)
a. The easiest way could be to segment the upstream
- Verify no uncorrectable forward error correction (FEC) codewords and “clean” plant
- Increase capacity without physical node splits or SG changes
a. Allocate more spectrum for high-speed data carriers
b. Take advantage of PowerBoost™ by eliminating
the typical 10% over-provisioning
- Eliminate Overhead
a. Split MAC domains to have fewer upstreams per domain
– More upstreams (USs) in a MAC domain creates more downstream (DS) MAP overhead at ~0.4 Mbps per US
b. Don’t make all the DSs primary-capable
c. Remove “stale” service flows
- Control abusers and denial of service (DoS) attacks
- Optimize CMTS efficiency
a. Load balancing
b. DOCSIS 3.1 graceful profile management and US/DS resiliency/partial mode
- Implement cache servers
a. Netflix, YouTube, and others may reduce video quality to save bandwidth
CMTS “Tweaks” Include
- Cable modem (CM) insertion interval – CM ranging opportunities
o Experiment with
(config-if) #cab insertion-interval auto 120 1000
- US range and data backoff
Cable upstream x range-backoff 3 6
o This can be experimented with in case CM transmissions collide at
init(r1)and have to back-off
- Throttle CM ranging
o Experiment with
[no] cable throttle-modem init-rate <1-1000> holdoff-time <5-100> flush-rate <100-1000>
o Suggested values: 32 CM/s; 45 sec; 300 CM/s
- Prioritize pre-registration traffic
o During high congestion, CMs that are trying to register could be “stalled” at init(d) since the default priority of this traffic is 0
(config) #cable qos pre-registration us-priority [0-7]
- Voice over IP (VoIP) call signaling insurance
o Utilize non real-time polling service (nRTPS) for call signaling as it provides non-contention request opportunities that will guarantee call signaling during high US congestion.
For more information visit our Cable Access website. Also, be sure to register for the upcoming Cisco Knowledge Network webinar on May 7, “How Is Your Broadband Holding Up,” to hear from Cisco expert John Chapman, Cisco Fellow and CTO, Cable Access Business Unit, and me, where we’ll share valuable information you can use today to quickly adapt your cable network.
Regarding abusers, you can use the “Subscriber Traffic Management” feature to temporarily cap speed and/or demote priority of excessive-traffic users
Agreed. Believe me, I have a 7 page paper (and growing) that was congested (pun intended) down to this 🙂
I think another option is to use the Subscriber Traffic Management feature to temporarily remove the 10% overhead on heavy users. This directly influences the available network capacity while still delivering the expected information rate.
I agree, but I mention, “Take advantage of PowerBoost™ by eliminating the typical 10% over-provisioning”.
I think this is a better option for end-customers to “ring the bell” for speed tests and actually “see” much higher than they pay for and the cable operator doesn’t have to over-provision 10% 24×7.
My personal opinion is powerboost can cause some odd behavior in particular with adaptive bandwidth algorithms. I think I’d rather deliver a more consistent service and only reduce bandwidth when necessary instead.
Good point. I’ve always thought about ABR applications and how they adapt over time and dependent on the rate available and screen you are using to watch. My thought was, Powerboost would be great for quicker buffering of content in the beginning 5-8 sec but once the rate goes down after the Powerboost, what happens with the ABR?
Maybe others have “real” results and data to share besides my speculations 🙂
Good stuff John. Always much better in person, but such is life these days.
I just did a 2-hr preso on this topic for the local SCTE chapter (Crystal Coast) here in NC. I’m sure they’ll have the recording up soon and i plan to condense the slides and send a PDF for them to post.
Comments are closed.