Effective Network Traffic Control in the Age of Always-On Video

It’s all about video these days. The growth of streaming video is giving mobile service providers headaches…as if making the move to 5G wasn’t tough enough! With millions more families now staying home and staying safe, mom and dad are video conferencing with colleagues, the kids are attending online classes, and that crazy uncle is streaming football highlights all day long. Yes, it’s a challenging time to be a service provider.

Already burdened by network bandwidth demands and increasing costs of providing 4G/LTE while transitioning to 5G, service providers are struggling with a tremendous bump in daily usage driven mostly by video. Streaming video causes large, continuous data flows, resulting in service slowdowns and unwelcome latency so everyone on the network suffers. In fact, the top 20 percent of mobile users account for a whopping 62 percent of mobile data traffic (Source: Cisco Annual Internet Report 2020, based on North American Tier 1-2 operators). And it’s not going to get any easier. Video will make up 79 percent of total mobile data traffic by 2022, according to the Cisco Visual Networking Index (mobile 2019).

While contending with customer expectations, increased speeds, more mobile connections, extra services, and a massive spike in video traffic, how are already taxed service providers going to accommodate this growth? And then there’s the cost to consider. Service providers must invest in revenue-driven solutions while at the same time decreasing Capital Expenditures (CapEx) and Operating Expenditures (OpEx). That’s a difficult balance to maintain.

We’ve rolled out a solution that’s already proving its worth to mobile service providers. Cisco Ultra Traffic Optimization (CUTO) is an enticing alternative to deploying costly spectrum and adding cellular radio gear. The software optimizes large traffic flows at congested cell sites to make the most of radio network performance where it’s truly critical – in densely populated, high-traffic areas.

 

What Causes Network Congestion?

A significant problem for mobile networks is elephant flow. These very large, continuous flows over a network link take up a disproportionate share of total bandwidth. Elephant flows are greedy, and they are asking cell sites for more bandwidth than they really need. Video is the primary contributor, along with large file downloads such as application and operating system updates.

For example, a single two-minute video consumes thousands of times more cell site capacity than two minutes spent browsing websites. It’s not surprising that all other consumer interactions such as web and application browsing, music, email, maps, and social media, make up only about one-third of traffic by volume.

Our intelligent, machine-learning software is designed to properly schedule the bandwidth by managing large flows and congested cell sites. Service providers can expect to see a 20% increase in throughput during RAN congestion, a 40% increase in cell site capacity, and a 20% reduction of RAN CapEx and OpEx. CUTO reacts on a second-by-second basis to real-time network congestion and optimizes all forms of traffic regardless of transport protocol (TCP, QUIC, UDP), encrypted or non-encrypted. Being vendor and technology agnostic, it runs either as part of a Cisco packet core or as a stand-alone on any type of network from 2G to 5G. Because CUTO does not perform DPI or modify the traffic in any way it is a highly scalable solution with low network touch integration.

 

How does Cisco Ultra Traffic Optimization work?

The scheduler software in each cell sorts traffic and places it into frequency and time slots for mobile device delivery. Streaming video saturates the scheduler, which causes latency, and the active user data sessions start backing off and slowing down. Eventually, the cell site capacity becomes exhausted. Because video uses far more concurrent resource blocks than it needs to, the scheduler over-allocates bandwidth to video users at the expense of the other 90 percent of subscribers. The result at the cell site is inefficiency and reduced capacity.

Cell site traffic load varies from second to second, so CUTO reacts to real-time congestion on a second-to-second basis to deliver proactive cross-traffic contention detection. The machine learning algorithms use the data delivery characteristics of the elephant flow data packets to determine if that flow is being delivered during moments of congestion. When experiencing network congestion, large video streams are flow-controlled packet by packet at the core, providing cycles to smaller flows while still preserving the same (or better) video quality at points of congestion. This removes the contention at the scheduler, frees up capacity, and greatly improves cell site performance.

CUTO looks at each IP flow uniquely and performs analysis to determine whether it’s a large flow and if it’s congested. Although it monitors all flows, it pays closer attention to large flows causing congestion. It’s a manageable device from the size and simplicity perspective as well. The 2-RU server is lightweight and runs right out of the box or can be a Virtual Network Function (VNF) application on a Network Functions Virtualization Infrastructure (NFVI) stack. It’s a rock-solid solution to network traffic congestion at an affordable price and can deliver millions in CapEx savings. And for the sake of simplicity and ease of use, network-wide software deployment is typically completed in about an hour.

 

Trials and Deployments

We validated deployments on networks across North America, Latin America, and Asia that showed it had the equivalent effect of adding spectrum. In a European service provider, we discovered that 55 percent of users were on sites that averaged 8 Mbps of increased throughput with a peak of 15.61 Mbps. Ensuring a hardware site capacity increase across all these cells would require a spectrum overlay of 10MHz.

In a separate trial, a 10-MHz eNodeB that was carrying 15Mbps now carries 22 Mbps. That’s the equivalent of adding a 4-MHz spectrum band in every cell in the network. Our trial in Latin America showed significant improvement in RAN KPI user throughput and we measured CUTO increasing bits per second/per user by 20 percent. It also demonstrated an average all-day service improvement of 20 percent, a peak improvement of 55 percent, and an overall cell capacity gain of 40 percent.

Spark New Zealand recently deployed CUTO software to prepare its network for 5G and to improve the overall mobile experience during the Rugby World Cup broadcasts. We know that elephant flows from streaming video can slow down or block other mobile traffic such as email and social media. Spark used our product to maintain a quality experience for rugby fans while also protecting other critical mobile services like voice and data.

 

Learn more

We would love to share our excitement about the power of CUTO with you. For more information, take a look at our webinar or read our white paper. And due to these challenging times, we are facing around the globe, as a special bonus, we’re offering a free 90-day Try & Buy of CUTO to owners of Cisco Ultra Packet Core. Reach out to your account manager to take advantage of this limited time offer.