Why is Low Latency DOCSIS Relevant to Gameplay and Why You Should Care

Gaming is a billion-dollar industry with billions of players, and they use broadband to get there. I’ve spent a lot of time discussing the importance of low latency DOCSIS over the last couple of years on my podcast and YouTube channel. Let’s dive into it a bit more. Low latency or ping is desirable to achieve smoother gameplay and a user experience without lagging. If the ping is around 1000 milliseconds or higher, it’s definitely going to result in players suddenly appearing or disappearing and overall gameplay running incredibly slow. Additionally, games where timing is crucial, like first person shooter (FPS) and racing games, suffer a greatly diminished experience with high latency. In short, time lags from high latency makes gaming far less enjoyable.[1]

Gaming is a huge marketing opportunity and will be a deciding factor among the more than 2.5 billion gamers around the world according to Newzoo and ICO partners. The gaming industry is expected to be worth over 90 billion dollars by 2020[2], with the average age for a video game player of 31 years[3]. If you have been thinking video gamers are 11-year olds who don’t make buying decisions, this may change your perspective. Looking at Table 1, one can see that the group of 46-60 year olds spend an average of 6.32 hours per week playing online video games.

Note the 26-35 year-old’s age group spends more time playing online games than any other age group. Gaming already has an active fan base too. In other words, fans watch other gamers game the same way people today watch live football games. This age group is also the age group making decisions on what Internet service provider (ISP) they will choose. If you are the ISP with high latency and your competitor has low latency, the decision to switch can be made by default over latency alone. 

Now is the opportunity to be competitive and have a very quantitative key performance indicator (KPI) — with some help from CableLabs and your vendors — low latency DOCSIS (LLD) is on its way.

CableLabs Updates MULPI for Low Latency DOCSIS

The good news is that CableLabs is here to help. In April 2019, CableLabs updated the MAC and Upper Layer Protocols Interface Specification (MULPI for short) to include low latency DOCSIS in section 7.7. You can download the specification here: http://bit.ly/low_latency

The objective of section 7.7 of the MULPI specification is to enable low latency DOCSIS to provide latency times as low as 1 ms without having to do anything other than upgrading your DOCSIS 3.1 CMTSs software/firmware along with your DOCSIS 3.1 modem firmware — pre-DOCSIS 3.1 equipment is not supported.

How is Low Latency DOCSIS Achieved?

Often times we may assume that delay in the DOCSIS network is created by transit time in the coax and optical fiber. This delay is actually very small adding 0.02 to 0.6 milliseconds to the total latency. The major delay in DOCSIS networks is caused by queueing delay and media acquisition. Queuing delay is caused by TCP acknowledgments in downstream and upstream data transmission. This creates a large round-trip delay for data, which is also dependent on the latency of both the downstream and upstream.

The second major latency element is media acquisition, which is the time spent primarily in the upstream while the CMTS allocates time for the modem to send upstream traffic.

CableLabs has focused on two key technologies in DOCSIS to address queuing delay and media acquisition as follows:

Active Queue Management

Dual queue and coupled AQM are more progressive versions of AQM. AQM has been around for some time and studies have been conducted to show its impressive improvement on voice over IP (VoIP) mean opinion scores. The main reason for this is that in many networks latency can easily reach over 150 milliseconds, which is the threshold for most people to notice an uncomfortable delay while talking on the phone. Latency becomes very noticeable at 300 milliseconds, which is often observed when watching live news when one news person is speaking to another in a foreign country over a satellite phone.

While we don’t desire our DOCSIS networks to feel like speaking to someone on a satellite phone, it can certainly happen under heavy congestion. This is another reason why we care about LLD. AQM is the first feature that enables a CMTS to make up to a 13x improvement on latency.[5]

Upstream Scheduling Improvements

Upstream scheduling improvements add up to another 80x improvement.[5]The upstream is notoriously challenging because every time a modem wants to transmit data it must first ask for permission, called a data REQuest or REQ. Then the modem must wait for a MAP message from the CMTS letting the modem know when it can transmit the data. Finally, when the timeslot arrives the modem can transmit its data. This is a latency laden process.

The above process can be improved by reducing the CMTS mini-slot time. That is the time between when a cable modem makes its data REQ and when the CMTS responds with a MAP grant to less than 1 millisecond. In current CMTSs this time could be multiple milliseconds, which would prevent the upstream from ever achieving low latency goals. This similarly impacts the downstream due to the TCP/ACK relationship.

The next major step that can be taken is proactive grant service (PGS). One can think about PGS similar to unsolicited grant service (UGS), which is commonly used for VoIP. In this case, when a user picks up the phone a dynamic quality of service (DQoS) flow is created specifically for the phone call. Since the CMTS knows that an upstream voice packet will be coming from the phone call every 20 milliseconds the CMTS does not need to wait for the REQ from the cable modem (CM). Instead the CMTS will automatically send a MAP grant to the CM in anticipation of the voice traffic. When the modem sees the MAP grant it can simply stuff the voice traffic into the grant with no delay added.

A similar concept is proposed for time sensitive traffic, such as gaming, Facetime, Skype, etc., in order to achieve low latency DOCSIS. The cable operator will need to define certain traffic types (such as those listed above) as “proactive” in order for them to be able to take advantage of PGS. Herein lies the challenge of what traffic becomes PGS and how is it done. 

Low Latency DOCSIS and the Future

We discussed the case for gaming at the introduction of this article, but traditional gaming is just the start. Let’s look at where we are today in Table 2 where early DOCSIS 3.0 equipment can have latency as high as 1 second under load. Even DOCSIS 3.0 equipment using buffer control under load will typically see latency of 100 milliseconds. This is very close to the 150 millisecond threshold for voice users and would be considered unusable for most gamers playing first person shooter (FPS) games.

High latency will become more pronounced as gaming with virtual reality (VR) adoption increases. A recent study by Vodafone and Ericsson in 2018 focused on the impact of latency and VR gaming that provided scientific evidence that latency had pronounced stress and discomfort for VR gamers.[7]

“The study found that the impact of poor network performance was particularly pronounced for VR gaming. Network lag in VR was three times more stressful for gamers compared to lag experienced while gaming on a PC. The study results provided valuable insights into the demands that future services, such as VR gaming, will place on network performance…”

VR gamers are the bleeding edge. This has more implications for non-gamer VR users who will become late adopters, but nonetheless VR users are on our networks. Further, we can look towards other uses of VR such as entertainment, healthcare, education, shopping, military, etc.

Low Latency DOCSIS and CMTS Vendors

To continue reading go to Broadband Library

Read about 5G vs Low Latency DOCSIS