Skip to main content
Sep 09 2013

Tracking Data Usage in a Multi-Device World

Shared data plans simplify life for mobile subscribers with multiple devices. But for operators, they add new complexity to the charging process.

Beyond counting voice minutes

Shared data plans are playing an important role in mobile networks’ transition from traditional voice to broadband data. They offer advantages for both operators and subscribers, as users turn to smartphones and tablets, and as multi-device ownership increases worldwide. But along with all their advantages, shared data plans present new challenges for a mobile operator’s charging system. The challenges begin with the different ways subscribers relate to the network, while using data services. Charging systems are no longer simply tracking the minutes used during individual voice calls, which have a clear beginning and end. Instead, they are tracking the amount of data used during always-on sessions, occurring across multiple users and devices. And, adding to the complexity, a single device can support multiple simultaneous sessions. When a charging system supports a shared data plan, it allocates data to diverse sessions by drawing on a shared monthly quota. Then, it works to maintain real-time accuracy about the remaining data balance, even while sessions are active. However, the processes used for data allocation include an inherent element of uncertainty about the data balance during the time a session is underway. And this uncertainty can lead to two types of problems:

  • Usage reporting that is not fully up to the minute.
  • Rejection of a new session, because the system is not aware, at the moment the session is requested, that a data balance is still available in the account.

This article focuses on helping mobile operators get the information needed to implement an efficient charging system for shared data plans. It takes a closer look at the processes of allocating data and tracking its usage across multiple users and devices. And it outlines key capabilities that can help address the biggest challenges. This information can be essential for meeting regulatory requirements, maintaining customer satisfaction and supporting new ways to monetize the network.

New relationships with the network

Shared data plans are valuable, because they recognize that subscribers are now relating to the network in new ways. Voice calls and messaging are in decline. Smartphones are proliferating, and broader deployment of LTE networks is allowing more subscribers to enjoy high performance for their wireless data applications. These trends are increasing the likelihood of subscribers owning multiple wireless connected devices. In this emerging multi-device world, subscribers need more flexible and cost-effective ways to use their data. Shared data plans meet these needs by allowing a specified data quota to be shared across multiple devices — and multiple users. That means subscribers no longer need a separate account for each device. And they can use any portion of their available data balance on any linked device, whether they want to watch videos, stream music, play video games or share files and photographs. By delivering this flexibility and convenience, mobile operators are gaining greater customer loyalty.[1] And they also have the opportunity to generate new revenues, for example, by making real-time offers on additional data.

Real-time tracking of simultaneous sessions

Along with the important benefits they offer, however, shared data plans also place new demands on a mobile operator’s charging system. One key goal of the system is to track how much of the data quota is still available for use. With shared data plans, the factors involved in tracking the balance can be quite complex, for the following reasons:

  • Smart devices offer always-on data sessions.
  • Multiple simultaneous sessions can occur on a single device.
  • Data usage patterns can be highly diverse across users, depending on the types of applications and services each user prefers.

Nevertheless, maintaining accurate real-time balance and usage information is important for a variety of reasons. The information is needed to meet regulatory requirements for usage reporting. It helps keep customers satisfied. And it can be an important element of new revenue-generating ventures, such as context-driven offers. When balance information is not up to date, the following issues commonly occur.

  • The mobile operator cannot deliver timely and accurate alerts, notifying subscribers when they have reached a particular threshold in their data usage — such as 50 percent or 80 percent.
  • A session request may be rejected, because the charging system is not aware, at that instant, that data is still available for use. This problem is commonly known as “starvation.”
  • Both these issues are most likely to occur when many data sessions are active simultaneously. To understand why this can affect real-time accuracy, it’s important to look more closely at how data is allocated to each device.

The data slice dilemma

When a mobile subscriber uses a device, it automatically sets in motion at least one session request. The network sends this request to the charging system, shown in Figure 1, which allocates a portion (or “slice”) of the available monthly data quota to the subscriber’s device. If the data session consumes all the allocated data, the network can request an additional slice. The charging system will then grant this request — if enough data remains in the monthly quota. When the subscriber’s session ends, any unused data from the allocated slice is returned. For example, Figure 2 shows the starting state of a shared account. The SurePay® charging system manages group level information which includes total monthly quota for a shared account, the quota used by the group so far, quota reserved for use by currently active sessions, any remaining quota and the corresponding monetary information. In this case, there’s a monthly quota of 1 Gb of which nothing is used or reserved yet.

As shown in Figure 3, the SurePay system initially allocates a data slice of 200 MB to Subscriber 1’s data session, and this amount is “reserved” from the group allowance. In subsequent updates of this subscriber’s session, the network can report how much data has been used and request an additional slice, if needed.

However, until the unused data in a slice is returned, the charging system does not know precisely how much of the allocated slice has been used. In other words, while any subscriber’s session is underway, there is a period of uncertainty about how much data will be returned to the shared monthly quota. And that gap in real-time knowledge also prevents a precise accounting of the data balance still available for use. When many data sessions are active simultaneously, the uncertainty about the remaining data balance increases, making starvation more likely and affecting the timeliness and accuracy of usage reporting. The trade-offs Both these common problems can be reasonably solved by allocating very, very small data slices. With this approach, the larger the number of simultaneous sessions, the smaller each slice needs to be. This solution is not entirely pragmatic, however, because smaller slices produce substantially greater signaling and processing load on both the packet data network gateway and the charging system. Sophisticated mechanisms for slice allocation are needed to retain the delicate balance between avoiding starvation and reducing signal load. In general, a variety of approaches can be used to determine the per session slice size for multiple simultaneous data sessions. For example, slices can be assigned as a portion of the full monthly quota, getting smaller as the month proceeds, but with a minimum and maximum applied. The size can be set according to static attributes, such as device type, access type or Cost of Service (COS) category. Or it can be determined with dynamic attributes, such as location and usage. A combination of these factors can also be used. However, these mechanisms do not usually offer deterministic guarantees relating to accuracy or starvation. More sophisticated mechanisms are needed, because they can incorporate strategies designed to provide such guarantees and, potentially, to allow timely changes in the slice allocation, if and when needed. Ideally, the slice allocated to a session will satisfy the needs of that session’s usage pattern, while reducing the overall signaling required for all the sessions in a shared account. If slices are allocated without considering how the slice gets used, a starvation situation may occur for a new session. To address this situation, some unused parts of data slices previously allocated to one or more other sessions could be reclaimed. This approach provides an alternative to rejecting the session. The potential costs of this approach depend on the size of the unknowns, such as how much of the allocated slice has been used up. The longer the session, the bigger the unknown. And the larger the group, the higher the potential cost may be, because the unknowns are multiplied by the number of users. Because the behavior of the system depends on how allocated data slices are used during a session, adaptive slice management can be valuable. It determines the slice size, based on some learned usage patterns, along with other system characteristics.

Charging capabilities to look for

To address the new challenges presented by shared data plans, mobile operators need to implement a charging system that can provide the following characteristics and capabilities: Flexibility Your system should be able to support a variety of individual and shared account types including hierarchical accounts. This flexibility allows customers to personalize their plans according to their unique needs. For example, devices in a shared account can be organized in a hierarchy. And shared accounts can also support line limits to provide the ability to set data usage constraints on individual users or devices. The charging system should also support prioritization, so data is allocated fairly when there is no way to satisfy all session requests. For example, a session for a health monitoring device would always get priority over a casual web browsing session. Innovative slice management techniques Charging systems, such as SurePay, are already focused on shared data plan challenges and are providing new techniques for allocating data. They offer new approaches to alleviating signaling issues, as well as diverse options for determining the optimal slice size. Robust overload control mechanisms To handle peaks and overloads, your system needs to provide robust controls and usage aggregation techniques. For example, members of a group can be spread across several OCS instances, while simultaneous activity within the group is given special attention to ensure data integrity. The growing importance of shared data plans The proliferation of smart devices is expected to continue. And more types of wireless connected devices are anticipated in the future, including smart meters, connected cars, devices for home monitoring and healthcare. As this transition from traditional voice to the multi-device world proceeds, it will be increasingly important to have a charging system that can support shared data plans efficiently. With the right charging capabilities, a mobile operator has better tools for keeping customers satisfied, meeting regulatory requirements and enabling new revenue opportunities. To contact the author or request additional information, please send email to networks.nokia_news@nokia.com.

Footnote

  1. [1]www.jdpower.com/content/press-release/lyD2xnz/2013-u-s-wireless-customer-care-full-service-performance-study-volume-1-and-the-2013-u-s-wireless-customer-care-non-contract-performance-study-volume-1.htm
About Pramod Koppol
Pramod Koppol currently serves as the CTO of the Payment, Policy & Charging business at Alcatel-Lucent. Prior to this, he served in a variety of technology leadership roles including head of Systems and Security research department in Bell Labs Research and head of product development in an Alcatel-Lucent venture addressing enterprise security. He joined Bell Labs after receiving a Ph.D. in Computer Science from North Carolina State University, Raleigh, in 1996.