The Urgent Need for Digital Media Subscriber Modeling

Thu, 01/27/2011 - 5:11am
Joseph Zeto, Ixia
Joe ZetoWith the advent of 4G wireless applications, the world is rapidly shrinking, drawing subscribers together in new and powerful ways. We are heading toward an always-on, always connected world. Engaging, media-rich mobile applications will become the norm and the productivity gains from this hyper-interactivity will be profound. Mobile devices such as smart phones, netbooks, and similar yet-to-be-designed tools will serve as on-ramps to the Internet and will allow anytime, anywhere access to critical information.User imagination and demand is easily outpacing network capacity and the journey to worldwide 4G LTE wireless has begun in earnest.

The transition from 3G to 4G wireless networksand the required interoperability with legacy technologies, however, will unleash a level of unprecedented complexity. Legacy technologies will need to seamlessly interact with newer technologies in order for service providers to deploy networks that not only attract subscribers, but limit maintenance and upkeep costs. To mitigate the risks, each network equipment supplier and service provider will need to use methodic planning, emulation, and network testing tools to validate performance and subscriber Quality of Experience (QoE). 4G networks will usher in increased bandwidth, complexity and lower operational costs—if deployed with systematic thought to end-to-end quality. 

Multimedia service modeling in Ixia’s Ixload test application

LTE networks feature very high subscriber data rates—up to 150 Mbps downlink—when using 2x2 MIMO antenna technologies. This speed will engender numerous higher bandwidth applications: video,data, and voice in unique combinations, requiring that network nodes handle high-bandwidth traffic with stringent jitter, latency and loss requirements. Triple-play and video rich traffic is essential for testing node functions that enforce quality of service (QoS), such as deep packet inspection (DPI).

Testing of next-gen converged networks requires independent testing of wireless and wired components as well as complete end-to-end testing. With multimedia and peer-to-peer applications accounting for the lion’s share of 3G/4G bandwidth usage, it is essential that pre-deployment testing utilize realistic traffic emulation. This ensures that networks are properly provisioned and that there are no surprises when users start to stress the live systems. Perception of the service quality is everything to the customer and is the lynchpin in preventing customer churn.

It is essential that wired and wireless components now be tested with the same types and scale of traffic. The iPhone is the perfect example of a high performance, multimedia-capable device and is the tip of the iceberg of what is to come. Networks, both wireless and wired, must forward and shape handheld-driven traffic so as to ensure balanced quality of experience (QoE) for all network users.

As a benchmark, people often refer to Moore’s law of integrated circuit (IC) density when speaking of computers. Moore’s law is a linear increase, with capacity doubling every two years. As smart wireless devices become the standard for Internet access, the increase demand will not be linear—it will be exponential. Such increases will require providers to understand a network’s ability to deliver services to millions of customers in a timely and consistent fashion. In order to accomplish this goal, providers must employ subscriber modeling in order to understand what their networks can handle in real-world situations. It’s not enough to just emulate users and applications —service providers must be able to test network reliability down to the behavior and experience of each individual subscriber.

Since 4G call durations will be much longer, with peaky data patterns (high while downloading a page, low when viewing a page), the carrier must maintain a higher number of concurrent calls with inconsistent bandwidth requirements. Subscriber modeling emulates real-world traffic patterns and data levels. Currently, it is estimated that in some providers only 3% of subscribers are using 40% of the available wireless bandwidth. As smart wireless devices become more and more prevalent, network providers must be able to test their networks with exponential amounts of data. They must be able to anticipate the large amounts of voice, video, P2P, gaming, and other data traffic that will deluge a wireless network. 

Traffic management requirements for mobile broadband applications

In addition to traditional network-based packet metrics such as latency, jitter, and packet loss, service providers need to use QoE testing to understand the user experience prior to network deployment. This requires the ability to correlate QoE statistics on an aggregated and per-user basis to determine if and when the network has not delivered to the service level agreement (SLA) for the subscriber’s services, and to understand the impact of one service on another while each competes for network resources.

Traffic shaping and QoS policies are critical to the integrity of these SLAs, and a growing and important trend in service provider network traffic shaping is the use of data packet inspection (DPI). DPI goes beyond the typical processing of header information and looks at the data within the packet to determine the type of traffic being processed. This allows a service provider to limit potential bottlenecks in its network by enacting more consistent QoS policies. Not only does DPI allow for better network control, but it prevents unauthorized users and service types, denial of service attacks and other malicious traffic disguised as legitimate applications, and allows providers to manage bandwidth-intensive traffic such as P2P. Subscriber modeling is the perfect means to emulate traffic patterns that include such types of traffic in order to test content-aware devices. 

Timelines overview

Subscriber modeling may be coupled with emulation testing, which can assess the network equipment for ability and resiliency and also verify that the network lives up to the expectations of a service provider’s customers. Subscriber modeling is best when it closely mimics the dynamic nature of consumer behavior.thereby allowing telecom equipment manufacturers (TEMs) and service providers to assess the performance of their service delivery. Providers can manipulate such scenarios through the following tasks:

• Define how users connect to their service provider network (for example, using IP, DHCP, PPP/L2P, IPSec, VLAN, 802.1P)
• Designate upstream/downstream bandwidth for each subscriber
• Specify which applications are used and they behave when traversing the network and interacting with each other
• Configure multiservice traffic distribution for each subscriber
• Model usage for each application with advanced time lines for traffic load profiling over an hour, a day, or a number of days; and
• Measure QoE per-application, including response latency, packet loss, jitter, MOS scores for VoIP, and MDI and MOS_V for video

The ability to do the above depends on which test system is deployed. Test systems also vary on their ability to test individual devices, subsystems, or entire networks from end to end. Moreover, some test systems are only capable of testing only a certain amount of traffic or perform only certain traffic protocol analysis. It is best to deploy a test system that can fully test the functionality and scale of network components and the evolved packet core (EPC). As carriers move toward new, untried 4G and LTE networks over the next few years, end-to-end testing will be critical for validating not only the functionality of the network, but its ability to deliver service that retains customer loyalty and bring unprecedented performance to wireless users.

About the Author:
Joseph Zeto serves as a technical marketing evangelist within Ixia’s marketing organization. He has over 17 years of experience in wireless and IP networking, both from the engineering and marketing sides.

Share this Story

You may login with either your assigned username or your e-mail address.
The password field is case sensitive.