The State of QoS 2017

Traditionally, QoS has only been addressable on the wired Ethernet front, in part due to the limitations of early Wi-Fi standards, which halved the potential throughput bandwidth to every connected Wi-Fi device the moment one additional device joined the Wi-Fi network.

Thankfully, the continued advancements in Wi-Fi standards have brought QoS options close to parity. The simplest QoS option, which is now being offered in both access points (AP) and routers, is the Wi-Fi Multimedia (WMM) standard, which is a subset of the IEEE 802.11e standard that targets distribution to devices based on their MAC addresses.

WMM automatically prioritizes network traffic in four categories, with two-way voice receiving top priority, followed by all types of video, and then lumping almost all other traffic into two areas: best effort delivery and background tasks.

The latter could include file downloads, meaning there’s a potential that HTTP video delivery, if packets are mislabeled as file downloads, could suffer in WMM.

Interestingly, a little-known feature of WMM, its Power Save certification, is now getting additional attention thanks to Apple’s recent iOS 10 release that features a low-power “dozing” option when the battery level reaches the 20% range.

“WMM-Power Save increases the efficiency and flexibility of data transmission,” a note from the Wi-Fi Alliance states. “Specifically, the client device can doze between packets to save power, while the access point buffers downlink frames. The application chooses the time to wake up and receive data packets to maximize power conservation without sacrificing Quality of Service.”

Cashing In on Caching

From a practical standpoint, the point on the network closest to the end user, at least the one that we can properly engineer, is the caching point.

A recent trial of the Open Cache specification, approved late last year by the Streaming Video Alliance (SVA), hopes to set a routing request standard for popular content caches that would be placed deeper in the ISP’s network (e.g., closer to the end user).

In some ways, the SVA work mirrors IETF work for HTTP caching, but the focus of the new trial—which includes Charter, MLB Advanced Media, Verizon, and Viacom, among others—is tightly narrowed to address OTT content.

Caching products that take advantage of the Open Cache specification should closely follow the current proof-of-concept trials, meaning that both enterprise and media and entertainment OTT delivery—including the large-viewing-audience live streams typical of MLBAM—should be able to take advantage of the specification’s benefits by mid- to late-2017.

Collaboration at the Core

As we move closer to the center of the network, two additional areas are aligning that may result in better QoS implementations.

First is the concept of CDN federation. This was a big topic a few years ago, but then died off when CDNs couldn’t agree on how to best charge one another for content traversing their respective networks and caching points.

In the pursuing years, a number of suggestions were raised on how to handle CDN federation, and the IETF took up the task of dealing with interconnections. We’ve recently covered this topic, including in this year’s CDN Buyers’ Guide.

Another area is the advancement of routers and managed switches in dealing with QoS settings.

For the QoS considerations close to the network core, we recommend an excellent Cisco book by Tim Szigeti, Christina Hattingh, Robert Barton, and Kenneth R. Briley, Jr., titled End-to-End QoS Network Design: Quality of Service for Rich-Media & Cloud Networks, 2nd Edition.

The book addresses just how much QoS matters these days, with the proliferation of types of video on an enterprise network:

In the early days of converged networking, applications fit cleanly into their respective media buckets: Applications were voice, video, or data— but not combinations of these. This was primarily because hardware was dedicated to application types. Today, however, because of the power and flexibility of hardware platforms—including laptops, tablets, and smartphones—many applications are multimedia hybrids, combining two or more of these media types.

In fact, multimedia streaming is just one of 12 classes of QoS-reliant applications, according to the recommendations listed in the Internet Engineering Task Force’s RFC 4594, an informational RFC. Application classes in RFC 4594 range from real-time interactive and VoIP telephony to broadcast video and multimedia conferencing (e.g., Webex, etc.).

“Conforming to it—in whole or in part—is the prerogative of the individual organization,” the authors write. “Such a complex model may be implemented in a gradual and phased approach.”

Managing to Differentiate

In the time since 2013, when the second edition of the Cisco QoS book appeared, a number of advances have taken place, both within the IETF and the Worldwide Web Consortium (W3C).

One of those is the automation of QoS best practices. In an email interview, Szigeti noted out that the recent automation of QoS processes benefits more than just network engineers: “Since publishing of this [2nd edition of the] book, we’ve programmed all of these recommendations into our EasyQoS application for APIC-EM, so that users wouldn’t have to read these 1000+ page books anymore.”

Another area is the potentially standardizing effect of WebRTC, a set of APIs that allow real-time communication protocols to be integrated into browsers and other JavaScript-programmable applications.

IETF is working on the protocol standardization while W3C is working on standardizing the APIs for several of the major browsers, allowing the browser to use the protocol to connect to back-end servers. The IETF version, titled Application Layer Protocol Negotiation for Web Real-Time Communications, has been waiting as a proposed standard since May 2016.

In the interim, an RFC has continued to progress forward that would allow for prioritization of streaming content on a packet-by-packet basis. Differentiated Services Code Point (DSCP) is a packet-marking approach that allows a form of QoS for browsers that support WebRTC applications.

This interim RFC is based on the RFC 2474 standard, first ratified in 1998. RFC submitters for this more recent DSCP for WebRTC, including Cisco and AT&T, see this version of QoS using packet marking as a way to offer default QoS for browsers in wireless networks, residential networks, and private wide area networks.

Within the IETFC draft specification, non-interactive video—what we could call streaming video—ranks third behind interactive video and audio, with two-way audio receiving the highest priority in expedited forwarding.

“For DiffServ QoS Treatment, RFC 4594 remains the guideline,” according to Szigeti, pointing to specific multimedia-streaming QoS guidance versus guidance for real-time or interactive video.

Should the draft of this newest RFC be ratified, it could potentially mean that managed switches and even home routers would have a way to easily prioritize the type of real-time video and audio content these devices receive.

Conclusion

QoS is a complex issue, not just for network architects but also for content publishers. After all, if content isn’t delivered in a timely manner at the proper bandwidth and resolution, all the network bandwidth in the world doesn’t matter.

While many of the QoS enhancements mentioned above require a network architect, there’s also a need to educate the consumer while at the same time to build infrastructures to address continually changing consumption patterns.

For instance, most consumers aren’t aware that their home gateways or routers often offer traffic-shaping technology on a per-device basis. Some consumer routers also include simple traffic identification (e.g., is video a high or low priority?), per-application priority, or even prioritization of specific ports beyond the HTTP port 80/8080 designation. And, while most consumer prioritization works on an IP basis, some of the more advanced home routers offer QoS assignment against a device’s permanent MAC address.

Consumer education is one key, but the story of 2017 will most likely be the use of real-time analytics services, not only for premium OTT content, but also for revamped enterprise video solutions.

The biggest challenge for QoS continues to be live events, whether corporate all-hands meetings or a national sporting event. Open caching holds promise, as do a number of draft IETF specifications that may make delivery of competing types of media content within a consumer’s home a much easier scenario.

This article appears in the March 2017 issue of Streaming Media magazine.

Latency Sucks! So Which Companies Are Creating a Solution?

It's one of the biggest challenges facing live video streaming today, and it's remained frustratingly stubborn. Explore the technical solutions that could finally bring latency down to the 1-second mark.

12 Jan 2017