Google’s QUIC made its debut a few years ago and it quickly became a headache for operators. Why? QUIC is an encryption-based protocol and many types of traffic become invisible. Operators struggled to deliver content and manage subscriber Quality of Experience (QoE).
Fast forward to the present and that conundrum is still the case making QUIC. One thing that has changed is that QUIC is now even more widely used!
Meta, Apple and Microsoft mimic QUIC
In 2018, QUIC was 20% of total mobile traffic. By the end of 2021, more than 75% of Meta’s (Facebook) internet traffic used mvfst. That is an experience-focused implementation from Meta of standardized IETF QUIC.
mvfst has shown significant improvements in several metrics. As per Meta, users experienced a 6% reduction in request errors, a 20% tail latency reduction, and a 5% reduction in response header size relative to HTTP/2. This also had a cascading effect on other metrics, indicating that peoples’ experiences were greatly enhanced.
The overall error count on video requests was reduced by 8%. The rate of video stalls was reduced by 20%. mvfst improved the video viewing experience, with an outsized impact on networks with relatively poorer conditions, especially those in emerging markets. Meta also deployed mvfst on Instagram for iOS and Android.
Apple has also embraced IETF QUIC. iOS 15 ships with HTTP/3 RTC — with QUIC transports — turns on by default.
Microsoft developed its own version and called it MsQUIC. It was open-sourced last year. Microsoft has been using MsQUIC to carry its Server Message Block (SMB) traffic and argued that SMB over QUIC is the future of distributed systems and that it enables use cases in edge computing and mobile devices — not possible to achieve over TCP.
Verizon Digital Media Services started building QUIC across its entire content delivery network to improve the network performance for customers.
Another example of QUIC going mainstream is its use as a 5G signaling protocol (3GPP 29.893). 5G service architectures are examining and identifying gaps and improvements.
How QUIC can benefit mobile operators
This trend means that it’s more important than ever for mobile operators to consider how QUIC affects their subscriber QoE. Google has already reported that QUIC promises to decrease the wait time for web search results by 8% on PCs and by 4% on phones. Similarly, QUIC also appears to lessen the buffering time for YouTube by 18% on PCs and 15% on mobile devices.
QUIC is here to stay but that said, some mobile operators still do not permit QUIC to be used and would actively block it as it would bypass other internal solutions for traffic capacity monitoring and management. An unforeseen side-effect of QUIC could be in mixed traffic (4G-5G-4G handoff) as QUIC’s inherent protocol changes make this process easier to manage from an application perspective.
On the topic of standardization, QUIC has been approved by the Internet Engineering Taskforce (IETF) to a certified standard — RFC 9000. But in truth, in the face of the performance improvements and heavily backed by the internet behemoths, IETF had little choice but to humbly accept QUIC.
Fact is, most mobile subscribers don’t know a thing about QUIC or any other protocols. They’ll just assume that this speed means their mobile operator has a faster network, making them less likely to churn.
QUIC blindsides mobile operators
So why is QUIC still a conundrum? As an encryption-based protocol, traffic is not visible at all to the mobile operator. Network providers cannot use traditional traffic management tools to control the cost to deliver the content, manage subscriber QoE and ultimately monetize their data.
This is particularly alarming when it comes to video content. Video accounts for approximately 70% of the total mobile internet traffic. Video represents more than 75% of the total QUIC traffic. Currently, mobile video in QUIC is using an Adaptive Bitrate (ABR) delivery mechanism where the server maintains several variants encoded at different qualities and bitrates. The client and the server negotiate the best version of the content to be delivered based on network conditions.
The concept seems ideal. The technology, however, introduces challenges to operators as ABR is an inherently ‘greedy’ protocol, consuming the highest bit rate that it can sustain unless video playback is interrupted. It also does not take collective QoE into account. It only looks at the individual video flow that it is serving. As such, QUIC in conjunction with ABR, debilitates the operator’s efforts to ease network congestion.
Google has certainly demonstrated their capability to reformulate internet standards — to their benefit. So, what steps can mobile operators take to manage incursions on their networks?
Start by reviewing their mobile data management strategy to identify and differentiate current and future traffic streams, such as QUIC, mvfst and MsQUIC.
Unfortunately, QUIC introduces additional challenges to those observed when managing other protocols built on top of TLS, such as HTTPS. The benefit of multiplexing multiple streams over a single connection comes with the added downside that it is impossible to differentiate between the different streams.
Recent Google QUIC version 50 (Q050) is already obfuscating the SNI. In parallel, work is currently underway in the TLS working group to encrypt the contents of the ClientHello in TLS 1.3 [TLS-ECH]. This would make SNI-based application identification impossible for on-path observation for IETF QUIC and other protocols that use TLS.
Mobile operators can combine transport layer optimization, encrypted traffic classification techniques and real-time RAN utilization. This provides a range of selective traffic management options for QUIC. These include content and user activity differentiation to create hand-crafted mobile data plans as well as QoE management that helps reduce congestion and the burden on the RAN.
Looking to the future
Content providers will continue to innovate proprietary protocols, driven by video and will use their commercial might to get these adopted and then retrospectively go to the standards bodies for approvals. In order to maintain visibility of traffic and relevance in the user experience, operators must have solutions for managing and optimizing UDP/QUIC.
In particular, UDP-Acceleration strategies are becoming indispensable. Operators also need to ensure that their vendor roadmaps include effective solutions for managing deep encryption.
Mobile operators can achieve reductions in the number of congested cells by 15%, facilitating fairness in the distribution of video bitrates (and therefore video quality) across subscribers sharing physical network resources.
It would be fair to say that nobody saw QUIC coming, but coming it sure is. By November 2022, approximately 95% of internet traffic could be encrypted and QUIC will be almost 47% of global internet traffic. So, will mobile operators bury their heads in the sand and mourn — or will they be ready for the fightback to take control of their network and their subscribers’ QoE?
