Online Charging with Diameter Gy:

Considerations for Accuracy and Reliability

An Industry Whitepaper

Contents
Executive Summary ................................... 1
Introduction to Online Charging .................... 2
3GPP Standards for Online Charging ............ 2
Considerations for Online Charging ................ 3
PCEFs and Intermediaries ......................... 3
Implications for Charging Accuracy and
Reliability ............................................ 5
Charging Accuracy .............................. 5
Charging Reliability ............................. 6
Support for Popular Use Cases ................... 6
Conclusion .............................................. 8
Requirements for an Online Charging Solution 9
Additional Resources .............................. 9

Executive Summary
Performing accurate, real-time metering of Layer-7 application
data traffic for prepaid charging use cases is a requirement in
todays telecommunications market. Accurate, real-time
charging is the foundation of many subscriber services, and is
necessary both to protect subscribers from bill-shock and to
protect network operators from revenue leakage.
According to the 3GPP standards, a direct, real-time interaction
between the PCEF and the OCS is required to ensure accurate
reporting and timeliness of the online charging mechanism. Many
systems claim to be 3GPP Gy-compliant, but lack one or both of
the functional requirements. These non-compliant solutions have
typically introduced a processing intermediary that breaks the
direct connection and adds many minutes of delay.
Non-compliance with the 3GPP online charging standards leads to
a host of accuracy and reliability challenges. In a non-compliant
architecture, attempts to address one shortcoming (e.g., lack of
accuracy) will make the other one worse (e.g., lower reliability).
The ultimate impacts on the operator include increased cost and
complexity, lower subscriber satisfaction, and revenue leakage.
At the surface (and marketing) level, it is often very difficult for
an evaluator to distinguish between compliant and noncompliant solutions. Therefore, an understanding of not only the
issues at hand but also the approaches used by various vendors is
required in order to make informed decisions and invest in a
solution that will actually work as needed.

Version 2.0

Online Charging with Diameter Gy: Considerations for Accuracy and Reliability

Introduction to Online Charging

Performing real-time metering of Layer-7 application data traffic is a pivotal requirement in todays
telecommunications markets. Accurate metering protects the revenue streams generated by data
services and ensures no bill shock for subscribers due to under- or over-counting.
To provide a standard model for deploying real-time metering solutions, the 3GPP has defined a set of
standards. Of particular relevance to real-time metering is the requirement that the measurement
component in the data path be connected directly and in real-time to the online charging system (OCS)
via Diameter Gy only in this manner can a communications service provider (CSP) avoid revenue
leakage.

3GPP Standards for Online Charging

Due to the importance of this subject, 3GPP standardization has defined online charging processes in
multiple standard specifications under the umbrella of the Policy and Charging Control (PCC)
framework as shown in Figure 1.
Subscription Profile
Repository
(SPR)

Sp
Online Charging
System
(OCS)

Policy and Charging Rules Function

(PCRF)
Sy

Gx

Gy

PCEF
Gz

Offline
Charging
System
(OFCS)

Figure 1 - 3GPP PCC Architecture (adapted from 3GPP TS 23.203 Release 11)

According to 3GPP TS 23.203, the PCEF is one logical entity and encompasses service data flow
detection, policy enforcement, and online and offline charging interactions. To ensure accuracy and
timeliness, 3GPP standards describe online charging mechanisms as real-time processes requiring a
direct interaction of the charging mechanism with data sessions and services. The charging mechanism
encompasses both the OCS and the PCEF.

Online Charging with Diameter Gy: Considerations for Accuracy and Reliability

Considerations for Online Charging

According to the 3GPP standards, a direct, real-time interaction between the PCEF and the OCS is
critical to ensure accurate reporting and timeliness of the online charging mechanism. Many systems
claim to be 3GPP Gy-compliant but fail in real-world deployments: how can this apparent contradiction
be reconciled?
There are two main ways in which a solution supplier will justify a claim of 3GPP online charging
compliance for a solution that actually fails to meet the standards:
1. Creative or partial interpretation and application of standards
2. Misrepresentation of components
The best way to understand these attempts is to critically examine them and compare them to the
actual meaning of the 3GPP standards for online charging.

PCEFs and Intermediaries

The figures below show three common online charging deployment scenarios, represented with a
simplified view of the network: Figure 2 shows a 3GPP-compliant deployment; Figure 3 shows a very
similar deployment, but this one includes an intermediary layer between the PCEF component and the
OCS, and therefore deviates from the standard; Figure 4 shows a second representation of the
deployment in Figure 3, that is functionally equivalent but visually confusing.

Figure 2 - 3GPP-compliant online charging deployment: direct, real-time connection

Online Charging with Diameter Gy: Considerations for Accuracy and Reliability

Figure 3 This deployment is non-compliant: the charging connection is not direct and is not real-time

Figure 4 - This deployment is non-compliant: the charging connection is still not direct and is still not realtime

Figure 2 is completely compliant with the 3GPP standard for an online charging deployment: the
measurement component (i.e., PCEF) has a direct, real-time connection via Gy to the online charging
system.

Online Charging with Diameter Gy: Considerations for Accuracy and Reliability
In the deployment shown in Figure 3, the measurement component (i.e., PCEF) lacks Gy. To obscure
this shortcoming, the PCEF vendor has added Gy to the subscriber management component, which now
sits between the measurement function and the online charging system. In this scenario, the PCEF
reports metering information periodically to the intermediary, which then processes the reports before
passing usage to the OCS.
Critically, the connection between the PCEF and the OCS is now neither direct, nor real-time: this
intermediate layer introduces significant delay (e.g., from a few minutes to as many as 30 minutes)
and guarantees billing problems of the kind discussed in the next section.
Figure 4 shows a deployment that is functionally equivalent to Figure 3; the only difference is that the
vendors misrepresents two completely separate components as being one. Without critically examining
this claim, an evaluator might conclude that the PCEF actually has a direct, real-time connection to
the OCS. However, strip away the superficial box drawn around the two components and they remain
functionally separate. Once again, the connection between the actual PCEF piece and the OCS is
nether direct, nor real-time.
If there was no difference in charging accuracy and reliability between these deployments, then noncompliance would not be of significant importance but that is not the case.

Implications for Charging Accuracy and Reliability

Charging Accuracy
In the indirect, interrupted connection, the timing between the PCEF node and intermediary node
introduces significant challenges with charging accuracy, especially when deployed in high-speed
networks (whether fixed access or mobile access).
In this architecture, the OCS grants limited user quota to the intermediary node. The intermediary
node continuously monitors the service usage from the PCEF to ensure that it does not exceed the
granted amount. To meet the real-time accuracy requirements of 3GPP TS 32.251 and TS 32.299, the
reporting frequency and thus the signaling load between the PCEF and the intermediary node are
unrealistically high. In contrast, the directly-integrated system shown in Figure 2 meets the 3GPP realtime requirements because metering and reporting are managed by a singular platform with on-wire
flow evaluation; hence, usage is reported immediately when the granted amount is reached.
Another serious issue with indirect integration is that accuracy errors grow quickly and proportionally
with the number of services that are commonly used in differentiated and tiered plans (precisely the
type of plan that is growing in popularity in prepaid markets). In this case, the direct charging method
offers superior predictability of the usage and revenue collection processes and hence better revenue
assurance.
For more advanced deployments that leverage rating groups (i.e., multiple services are combined into
a group) and service identifiers, the indirect integration method is even less effective. Whether or not
the intermediate node is used to aggregate services into rating groups, the existence of an additional
information transfer point amplifies error. If there are three services with 10% error, at the rating
group level that error can sum to 30% inaccuracy. With direct integration there is no accumulation of
error, because usage from all services and rating groups is monitored using the same charging process
and as a result the likelihood of usage overshoot is minimized.

Online Charging with Diameter Gy: Considerations for Accuracy and Reliability
Short-lived sessions pose unique challenges to the indirect integration method because the session
might terminate before the intermediary node initiates the connection with the OCS and is therefore
unable to provide the right grant amount to the PCEF. The problem becomes much worse when PCRF
signaling is needed in advance to approve a connection. As a result, with indirect and interrupted
charging architectures, the network operator will either suffer from significant revenue leakage or poor
subscriber quality of experience for services built around applications with short-lived sessions.
Unfortunately for the network operator, short-lived sessions are extremely common today across a
wide range of subscriber-centric and generalized network-wide use cases. 1 In such cases, direct
integration methods are recommended to avoid revenue leakage or degraded QoE.
Finally, in deployments where data traffic flows on asymmetric routes 2, the indirect architecture faces
even more challenges in trying to ensure quota alignment between the PCEF and the intermediary
node, resulting in more revenue leakage. This is less of a challenge in systems that directly integrate
with the OCS because quota alignment processing is minimal.

Charging Reliability
In addition to challenges with charging accuracy, systems that integrate indirectly with the OCS suffer
from significant challenges when it comes to reliability; these challenges are especially problematic
when accuracy requirements are stringent because the signaling load between the PCEF and the
intermediary node is high.
To ensure reliable transmission of charging reports between the PCEF and the intermediary node,
designers must consider trade-offs when choosing between using either a reliable (e.g., TCP) or
unreliable (e.g., UDP) packet transport protocol: with a reliable transport protocol, packet
retransmissions can result in network delays that increase the likelihood of revenue leakage due to
overshoot; with an unreliable transport protocol the delay is minimized but packet (i.e., reporting)
losses result in revenue leakage due to the loss of charging information for a measurement period.
In short, when using an indirect online charging architecture there is no guaranteed way to eliminate
revenue leakage when it comes to transporting charging information between the PCEF and the
intermediary node, which is why the 3GPP standard is strict about maintaining a direct, real-time
connection between the PCEF and the OCS. The higher the accuracy requirements, the higher the
signaling load and hence the higher the likelihood of either retransmissions or packet losses resulting in
even higher revenue leakage. These issues are insignificant when it comes to systems that directly
integrate with the OCS.
An additional downside of the indirect architecture is further overhead when connecting to a Network
Management System (NMS), introducing higher cost and more complexity due to having more nodes.

Support for Popular Use Cases

Network operators are increasingly introducing subscriber services that rely on accurate, reliable online
charging solutions. Table 1 examines some of these popular use cases and explains the potential for
revenue leakage using the two billing architectures examined in this paper.

Examples range from instant messaging and weather applications to M2M devices like smart meters and sensors that exchange
small data bursts short sessions are prevalent, and theyre here to stay
2
You can learn more about routing asymmetry in general, and its implications for network policy control, in the Sandvine
whitepaper Applying Network Policy Control to Asymmetric Traffic: Considerations and Solutions

Online Charging with Diameter Gy: Considerations for Accuracy and Reliability
Table 1 - The impact of online charging architecture on revenue leakage for popular use cases

Revenue Leakage
PCEFOCS:

PCEFOCS:

Direct
Real-Time

Not Direct
Not Real-Time

Basic Data Plans

None or minimal

Medium

Differentiated or
Tiered Plans

None or minimal

Medium to High

Roaming

None or minimal

High

Tethering

None or minimal

Medium to High

Blackberry and
Smartphone Plans

None or minimal

Medium to High

Use Case

Shared/Family
Plans

None or minimal

Additional Notes

The revenue leakage in the indirect integration case increases

as a function of the connection link speed

Subscribers usage is not accounted for and operators still need

to pay data transit costs to their transit partners
More PDP contexts are created in this case and hence more
accuracy and reliability issues between the PCEF and the
intermediary node
More PDP contexts are created in this case and hence more
accuracy and reliability issues between the PCEF and the
intermediary node

Very High

Inaccuracies are proportional to the average number of devices

per user, number of PDP contexts per device, and number of
services and rating groups that need to be tracked; if
subscribers are roaming, then revenue leakage is much worse.

Enterprise (high
value subscribers)
group accounts

None or minimal

Very High

Indirect integration architecture performs poorly when it comes

to supporting modern enterprise packages (e.g., enforcing usage
caps or speed limiting for small businesses, enforcing usage caps
on a per department basis, sending timely notifications to
business users during roaming about their roaming charges)

Machine-to-Machine

None or minimal

Very High

Short sessions are a big challenge to indirect architectures and

hence suffer from significant challenges in M2M deployments

Prepaid Plans

None or minimal

High

Zero-Rating

None or minimal

Medium to High

Time of Day
Bundles

None or minimal

Medium

Public Promotions
and Loyalty
Campaigns

None or minimal

Low

Sponsored
Connectivity

None or minimal

Medium

Revenue leakage can be a significant percentage in cases of

small quota with prepaid packages
Under-counting zero-rated bytes over-charges subscribers; overcounting zero-rated usage causes revenue leakage.
Time-of-day bundles are sensitive to the accuracy of usage
charging before and after the time when the charging rate
changes, it is important to ensure a clean and accurate split
between charging periods (much easier with direct OCS
integration due to the timeliness); with indirect architectures,
delays and time synchronization between the PCEF and the
intermediary node can result in revenue leakage in such cases
Depending on the number of users enrolled in a promotion,
usage can be inaccurately attributed to base plans rather than
to the promotional bundles, resulting in a confusing experience.
Due to delays applying zero-rating on sponsored content, the
subscriber might be penalized for promotional content or the
operator might lose revenue when sponsored connectivity is no
longer valid, resulting is zero-rating billable usage

Online Charging with Diameter Gy: Considerations for Accuracy and Reliability

Conclusion
To provide a standard model for deploying real-time metering solutions, the 3GPP has defined a set of
standards. Of particular relevance to real-time metering is the requirement that the measurement
component in the data path be connected directly and in real-time to the OCS via Diameter Gy.
Many systems claim to be 3GPP Gy-compliant but fail in real-world deployments because the claims are
false. Figure 5 shows three diagrams that represent two charging architectures:

The left-most architecture is compliant with 3GPP online charging standards: the connection
between the measurement component in the data path and the OCS is direct and is real-time
The middle figure is not compliant: the introduction of an intermediate processing node breaks
the direct connection and breaks the real-time nature of the usage reports
The right-most figure has an identical architecture to the middle figure, but the diagram has a
superficial box misrepresenting two separate components as one: this box does not rectify the
indirect, none real-time nature of the architecture

Figure 5 - Three diagrams representing two charging architectures: only the left-most diagram is compliant
with 3GPP standards for online charging

From the communication service providers perspective, failure to adhere to the strict 3GPP standards
for online charging results in revenue leakage and unreliable billing for all online charging use cases.
The degree of revenue leakage varies by use case based upon a handful of factors, so each must be
considered separately.
The only way for a CSP to achieve minimum revenue leakage and maximum reliability is by adhering to
the 3GPP requirements.

Online Charging with Diameter Gy: Considerations for Accuracy and Reliability

Requirements for an Online Charging Solution

The requirements for an accurate, reliable, 3GPP-compliant online charging solution are quite
straightforward:
Consideration

Requirement

Usage Measurement

The measurement component (e.g., PCEF,

TDF) must measure data usage in real-time

Connection between
Measurement
Component and
Online Charging
System

Connection between the measurement

component and the online charging system
must be direct (i.e., no intermediate
processing node)
Connection between the measurement
component and the online charging system
must be real-time (i.e., no interruption,
aggregation, etc.)

Explanation
To decrement from prepaid
quota in real-time, the data
usage must be measured in realtime (as opposed to later via a
records system)
This is the only way to ensure
accuracy and reliability in realworld networks
This is the only way to ensure
accuracy and reliability in realworld networks

Additional Resources
In addition to the 3GPP resources cited in this document, please consider reading the Sandvine
technology showcase: Standards-Compliant Online Charging, available on www.sandvine.com.

Headquarters
Sandvine Incorporated ULC
Waterloo, Ontario Canada
Phone: +1 519 880 2600
Email: sales@sandvine.com

European Offices
Sandvine Limited
Basingstoke, UK
Phone: +44 0 1256 698021
Email: sales@sandvine.co.uk

Copyright 2015 Sandvine

Incorporated ULC. Sandvine and
the Sandvine logo are registered
trademarks of Sandvine Incorporated
ULC. All rights reserved.