Smart & Secure Blog

Quality of Experience – How machine learning, net neutrality, and 5G relate to QoE

Quality of Experience – How machine learning, net neutrality, and 5G relate to QoE

Around the world, an enormous amount of money is being spent on bandwidth and network equipment as communications service providers (CSPs) strive to deliver optimal Quality of Experience (QoE) across a range of existing and planned services.

During 2018, in the US alone, communication service providers’ (CSP) broadband CAPEX rose more than $2.2B. Looking forward, McKinsey predicts that 5G will cause network TCO to grow by 60% between 2020 and 2025.

Not all these numbers reflect pure bandwidth costs – but, clearly, an enormous amount of money is being spent on bandwidth and network equipment as CSPs work to deliver optimal Quality of Experience (QoE) across a range of existing and planned services.

We surveyed CSPs about QoE, and their thoughts and our analysis appear in the Driving QoE Telco Smart Trends Report that was just published. (More details about this report, and how to get it, are available below.)

 

Earlier this year, I had a chance to interview Mark Shteiman, the VP Product Management here at Allot. Mark is a telecom industry veteran with rich multi-disciplinary knowledge and an impressive understanding of the B2C and B2B markets and we posted our conversation online in case you’d like to watch.

The June 5, 2019 interview with Mark Shteiman (L) and Michael Schachter (R).

 

Written below, transcribed for easier reading, are some of the most fascinating parts of our conversation.

 

MICHAEL: It’s pretty clear today that telecom companies, and CSPs in general, need to extract more value out of their networks and increase customer satisfaction. How can Allot help with this twin challenge?

MARK: Let’s first address the major trends that the CSPs experience. First of all, there is network capacity. As you know, network capacity is growing 40% year-over-year, and 70 percent of that growth is utilized by video and interactive services like gaming. So, basically, the customer expectations are continuously going up while, on the other hand, the budgets of the operators, both CAPEX and OPEX, are going down. That’s due to either stable or declining ARPU.

So, if the operators had the budgets, then improving customer satisfaction would be really easy — you throw in more hardware or more virtual resources. But, doing that (without big budget), while maintaining and optimizing your existing network, is really hard and that’s at the core of the Allot offering.

Basically, we allow operators to utilize their network up to the limit while, at the same time, protect the quality of experience of each of the subscribers and each of the applications.

 

MICHAEL: So QoE, the customer experience, is clearly a key challenge or driver for CSPs… What are the core technologies that Allot employs to help achieve this goal?

MARK: So, basically, all our technology is based on two key pillars. One is analytics, or closed-loop analytics. The second one is machine learning.

We believe that, these days, in order to properly and efficiently operate the network, it’s not enough to monitor only network APIs, such as bandwidth, RTT, retransmission… You have to really monitor the QoE, and QoE is really hard to analyze and understand because it’s basically a function of a lot of different parameters. It’s a function of the device. It’s a function of location. It’s a function of environmental conditions. It’s a function of network utilization, or that of another entity. So basically, the solution has to analyze in real-time and dynamically understand all these parameters and automatically implement all the required changes in the network.

Another challenge that exists in the network today is related to encryption. The majority of traffic is encrypted. So, it’s really hard to understand the actual QoE of the encrypted application. That’s why machine learning is the key enabler of everything we’re doing. We are investing in different types of machine learning… supervised learning, unsupervised learning, deep learning, and many others. Basically, we are building the whole analytics portfolio on top of machine learning.

 

MICHAEL: So, using machine learning and this deep analytics, you’re able to learn more about what’s transpiring in the network and what the users are experiencing. But, how do you then turn that into improved QoE and better utilization?

MARK: We gather a lot of historic information because we sit inside the network, and we gather a lot of real-time information, too. Then, when there’s an issue in the network, we are actually able to detect the degradation of the QoE of that, or another service, at that or another location.

Once we are detecting a degradation of QoE, we are able to implement different techniques in order to fix the issue and improve the QoE, or restore the QoE to the required level. These techniques include prioritization of mission-critical applications, or limiting traffic of less essential background services, optimizing the RAN scheduler or video optimization, prioritizing different service plans, etc. All of these different techniques can be deployed in an automatic way into the network to resolve the issue. By that, we’re improving the QoE end-to-end, both from the detection up to actual mitigation in the network.

 

MICHAEL: But Mark, if the CSP can now prioritize certain traffic at the expense of other traffic, doesn’t that contradict net neutrality rules, which are meant to guarantee an even playing field for all applications and services?

MARK: Operators who don’t deploy the Allot solution will drop traffic. But, which traffic they will drop? Any traffic — they don’t really differentiate. They can drop mission-critical traffic, or drop traffic of one user in favor of another. By that, they actual introduce unfair treatment.

With Allot in the network, they’re able to see and drop only traffic that really can be dropped. So, we can ensure that the mission-critical services will continue to function. If user traffic has to be dropped, we ensure that all users are treated in a fair way, basically dropping traffic in a proportional way, maintaining fairness on the network, allowing the operators to meet regulatory requirements.

 

MICHAEL: So, in essence, you’re saying that the Allot solution will actually help the service providers more fairly control and prioritize the traffic in a way that’s fair for everybody. But, doesn’t the move to 5G with service-based networking guaranteeing end-to-end quality in some way make this kind of a solution obsolete?

MARK: Definitely not. It’s true that 5G architecture is built around service delivery, actually defining new mechanisms, called network slicing, which are supposed to ensure end-to-end service delivery and quality. However, the amount or the number of network slices defined in a network will be limited in order to avoid a resource scarcity issue. So the limited number of slices will be defined for a group of applications with similar QoE requirements, or for a group of users with similar requirements. Enterprise, for example.

However, somebody will still have to ensure that, within this slice, between different applications, this specific application QoE is preserved. The same is true also for prioritization and resource allocation between the slices. It’s true that 5G comes with much higher bandwidth and much higher capacity. However, it’s also introducing new types of applications, and these applications will utilize the bandwidth. We’re talking about V2V vehicle-to-vehicle, assisted driving, massive IoT, augmented reality, just to mention a few new services.

So, at the end of the day, the capacity provided by the network will be utilized. At some point of time, at some location, there will be congestion. When that happens, there should be somebody in the network to ensure that resource allocation and prioritization is handled in the most efficient and proper way, and that somebody is Allot.

 

A deeper understanding

To better understand CSP attitudes towards QoE – what they measure, how they augment it, what threatens it, and where it is most important to their customers – we surveyed 106 CSPs around the world.

Their answers and valuable insights are available in the recently published Driving QoE Telco Smart Trends Report.

The report provides critical insights into how CSPs approach QoE today and how they view it tomorrow, as the industry transitions to 5G.

  • Discover industry trends and how key players in the telecom industry rate QoE’s centrality to their business.
  • Understand the impact of intelligent congestion management on assuring QoE.
  • Learn how proper implementation can save tens of millions of dollars each year.

 

Get the full Driving QoE report to learn more. 

 

Contact sales

Contact Sales

CONTACT SALES

Discover the best solutions for your organization

You’re all set!

We look forward to meeting with you on Monday, June 28 @ 14:00 EST. The meeting details will be sent to your mail box in a few seconds.

For a deep dive into Allot’s SMB solutions, we’d like to offer you a free copy of our position paper
Security for SMBs: Threats and Opportunities on the Rise.

Magazine Get your e-book »