Ir 3444
Ir 3444
Ir 3444
Copyright Notice
Copyright © 2008 GSM Association
GSM™ and the GSM Logo™ are registered and the property of the GSM Association.
Document History
CR
Version Date Brief Description
number
0.01 -
22.2.2000 Initial drafts & first issue
1.0
1.0.1 14.3.2000 Modifications after GPRSWP#8. Submitted to IREG#38 for approval.
2.0.0 15.3.2000 IREG 38 approval
3.0.0 28.4.2000 Approved at Plenary 43. PL Doc 35/00
CR from GPRS Doc 51/00 incorporated
3.1.0 5.9.2000 GPRS DNS Usage Guidelines incorporated as annex A
Approved at Plenary 44
SCR 003 to IR.34 Incorporated - Changes related to Quality of
3.2.0 19.10.2001
Service - SCR IR.34(v3.2.0)
CRs from IREG Doc 035/02Rev1, 036/02Rev1, 039/02 and 040/02
3.3.0 20.05.2002
to IR.34 Incorporated
IREG#44 Docs 041/03, 016/03Rev1, 050/03 and 033/03
3.4.0 28.01.2003
incorporated
3.5.0 20.10.2003 IREG#45 Docs 013/03, 015/03, 016/03 and 017/03 incorporated
3.5.1 07.01.2004 IREG Doc 46_011 incorporated
August
3.5.2 IREG Docs 047_012_rev2 and 047_018 incorporated
2004
February
3.6 Packet Doc 025_006 incorporated
2006
Removal of DNS specific information (which can now all be found in
3.7 April 2006
GSMA PRD IR.67). The references have also been updated.
November
4.0 Major Revision to include IPX information
2006
Restructuring to improve readability for non-GSMA parties. New
January
4.1 Architectural Description section. New GRX-IPX connectivity section
2007
and community attribute rules
October Major Revision to QoS information and minor modifications to IPX
4.2
2007 proxy information.
NCR
4.3 April 2008 Packet Doc 033_004 incorporated (Jitter requirements).
001
NCR Packet Doc 035_013r1 incorporated (Extended BGP communities
4.4 June 2008
002 and Hot potato routing).
Other Information
Type Description
Document Owner GSMA IREG Packet
Revision Control As required
Document editor/company Marko Onikki, TeliaSonera Finland
Feedback
This document is intended for use by the members of GSMA. It is our intention to provide a
quality product for your use. If you find any errors or omissions, please contact us with your
comments. You may notify us at mailto:prd@gsm.org. Your comments or suggestions are
always welcome.
Table of Contents
1 GENERAL........................................................................................................ 6
1.1 PURPOSE ....................................................................................................................... 6
1.2 BACKGROUND ................................................................................................................ 6
1.3 ABOUT THIS DOCUMENT ................................................................................................. 6
2 DOCUMENT SCOPE ....................................................................................... 7
2.1 IN SCOPE........................................................................................................................ 7
2.2 OUT OF SCOPE............................................................................................................... 7
3 DEFINITIONS, ABBREVIATIONS AND SYMBOLS........................................ 7
3.1 DEFINITIONS AND ABBREVIATIONS ................................................................................... 7
3.2 SYMBOLS ....................................................................................................................... 9
4 INTRODUCTION.............................................................................................. 9
4.1 THE NEED FOR IP INTERCONNECT ................................................................................... 9
4.2 GRX ............................................................................................................................ 10
4.3 IPX .............................................................................................................................. 10
5 INTER-SERVICE PROVIDER IP BACKBONE ARCHITECTURE ................. 10
5.1 SERVICE PROVIDER TO INTER-SERVICE PROVIDER IP BACKBONE CONNECTION ............. 11
5.2 GRX ARCHITECTURE .................................................................................................... 11
5.3 IPX ARCHITECTURE ...................................................................................................... 12
5.4 INTERCONNECT FUNCTIONS OF THE IPX........................................................................ 13
5.4.1 Transport-Only Connectivity Option................................................................... 13
5.4.2 Bilateral Service Transit Connectivity Option..................................................... 13
5.4.3 Multilateral Service Hub Connectivity Option..................................................... 13
5.5 IPX PROXY SERVICES .................................................................................................. 13
5.6 TYPES OF SERVICE PROVIDER AND INTERCONNECTIVITY ALLOWED ................................ 14
5.6.1 MNO-G .............................................................................................................. 14
5.6.2 MNO-I ................................................................................................................ 14
5.6.3 NGNO ................................................................................................................ 14
5.6.4 Permitted Interconnectivity ................................................................................ 14
5.7 TYPES OF IP BACKBONE PROVIDER .............................................................................. 15
5.7.1 GRX Provider..................................................................................................... 15
5.7.2 IPX Provider....................................................................................................... 15
6 REQUIREMENTS OF THE INTER-SERVICE PROVIDER IP BACKBONE .. 15
6.1 GRX PROVIDER REQUIREMENT .................................................................................... 15
6.2 IPX PROVIDER REQUIREMENTS .................................................................................... 16
6.3 CONNECTIONS BETWEEN IP BACKBONE PROVIDER AND SERVICE PROVIDER .................. 16
6.4 PEERING INTERFACE .................................................................................................... 16
6.5 TECHNICAL SPECIFICATION OF THE INTER-SERVICE PROVIDER IP BACKBONE ................ 17
6.5.1 IP Routing .......................................................................................................... 17
6.5.2 BGP-4 Advertisement Rules.............................................................................. 18
6.5.3 BGP Extended Community Attributes................................................................ 19
6.5.4 IP Addressing .................................................................................................... 23
6.5.5 DNS ................................................................................................................... 23
6.5.6 Security and Screening...................................................................................... 23
6.5.7 QoS.................................................................................................................... 24
6.5.8 Generic Proxy Requirements............................................................................. 24
7 TECHNICAL REQUIREMENTS FOR SERVICE PROVIDERS...................... 25
7.1 GENERAL SERVICE PROVIDER REQUIREMENTS ............................................................. 25
7.1.1 Service Provider IP Routing............................................................................... 25
7.1.2 Service Provider IP Addressing ......................................................................... 25
IR.34 Version 4.4 Page 4 of 52
GSM Association
Official Document: IR.34 Unrestricted
1 GENERAL
1.1 Purpose
The purpose of this document is to provide guidelines and technical information on how Inter-
Service Provider IP Backbone networks are set-up and how Service Providers will connect to
the Inter-Service Provider IP Backbone.
This document provides a clear distinction between GRX and IPX (evolved GRX) networks
where applicable.
This document also defines high level security requirements for the Inter-Service Provider IP
network. Detailed complementary requirements can be found in the “Inter-Operator IP
backbone Security Requirements For Service Providers and Inter-operator IP Backbone
Providers” IR.77 [19].
1.2 Background
The Inter-Service Provider IP Backbone network was originally created to carry GTP-tunnels
(GPRS Tunneling Protocol) via the Gp interface between the GPRS Support Nodes (GSNs) in
different GSM Operators ie data roaming. The Gp interface allowed mobile end-users to make
use of the GPRS/3G services of their home network while roaming in a visited network. Later,
MMS interworking and WLAN (authentication) data roaming has been added to the services
supported in Inter-Service Provider IP Backbone. This Inter-Service Provider IP Backbone is in
fact an Inter-PLMN IP Backbone and is termed the GRX. The GRX model is used to
interconnect in excess of 300 networks and has proven highly successful.
With the development of IP-based services, interworking of such services has become an
industry wide challenge. The GRX model is applicable as an IP interworking solution; however
the GRX specification does not meet all the requirements. It has been recognised that by adding
interworking specific functionality to the GRX model and offering it to the industry, a common
interconnect platform could be established for IP interworking. The enhanced GRX is called an
IPX and is designed to support a variety of types of Service Providers in a secure and business
sustainable way.
The core enhancements to the GRX are end-to-end Quality of Service and the introduction of
the IPX Proxy which facilitates interconnect cascade billing and multi-lateral interconnect
agreements.
1.3 About this Document
The document provides a brief introduction to the requirement for IP interworking and the IPX. It
covers the background to the forerunner of the IPX, the GRX.
The technical architectures of both the GRX and IPX are described followed by the technical
implementation guidelines for IPX and GRX Providers and connecting Service Providers.
Technical guidelines for Security, Quality of Service and Traffic applications are also given.
Appendices provide details on known issues in the Inter-Service Provider IP Backbone and on
the requirements for IPX proxies.
Note: In this document all references to the IP Inter-Service Provider IP Backbone Network (or
abbreviated to IP Backbone Network) shall include both IPX and GRX where generic
requirements exist. However specific requirements to IPX and GRX shall use each term
independently. In addition, the term “evolution to an IPX network” shall not in any way suggest
or mandate the GRX is either invalid or that it shall be replaced by an IPX.
2 DOCUMENT SCOPE
2.1 In scope
An Inter-Service Provider IP Backbone network architecture which connects Mobile Network
Operators (MNOs), Fixed Network Operators (FNOs) Internet Service Providers (ISPs) and
Application Service Providers (ASPs), from here on in referred to collectively as "Service
Providers". Where there is specific reference to an Service Provider type they shall be directly
referred to in each case.
Technical guidance to Service Providers for connecting their IP based networks and services
together to achieve roaming and/or inter-working services between them.
Recommendations for IP addressing. (Applies to inter- and intra-Service Provider nodes only.)
Host name recommendations remain within the scope of the present document.
Service Provider Mobile, fixed operator or other type of Operator connecting to Inter-
Service Provider IP Backbone for roaming and/or interworking
purposes
Whitelist A list supplied by a Service Provider of interworking or roaming
partners with whom connection is allowed
Hot Potato A term typically used for routing decision where a party is handing
over its traffic to a peering partner as quick as possible or at the
nearest in terms of delay peering point. I.E. when two SPs are on
different continents and behind two different IPXs, Hot Potato
means that IPX1 hands over SP1's traffic to IPX2 at
the Peering point nearest to the SP1's location. More information is
given in IETF RFC documentation [24]
Cold Potato A term typically used for routing decision where a party keeps its
traffic on its network for as long as possible and handing over its
traffic to a peering partner at the farthest in terms of delay peering
point. I.E. when two SPs are on different continents and behind two
different IPXs, Cold Potato means that IPX1 hands over SP1's
traffic to IPX2 at the Peering place farthest to the SP1's location.
More information is given in IETF RFC documentation [24]
3.2 Symbols
For the purposes of the present document, the following symbols apply [3]:
4 INTRODUCTION
4.2 GRX
The GRX Network was first established in 2000 for the purpose of Mobile GPRS roaming and
only MNOs were allowed to connect to it. Since then, other services have been added such as
UMTS roaming, MMS interworking and WLAN (authentication) data roaming.
The GRX provides connectivity based upon best-effort between GSM and 3G Mobile Network
Operators (end-to-end) whenever bilaterally agreed between those Operators. It includes the
agreement of the IP Backbone Provider to carry diagnostic protocols, for example ICMP (Ping).
4.3 IPX
Building on the features of the GRX, the IPX also is able to support the following:
• Connectivity between any type of Service Provider
• End-to-end QoS for roaming and interworking
• Any IP services on a bilateral basis with end-to-end QoS and interconnect charging
An IPX may also use the service-aware functionality of the IPX Proxies to support:
• Further interconnect charging models such as Service-Based Charging in addition to the
volume-based model of GRX
• Inter-operable interworking for specified IP services
• Multilateral interworking support for these specified services over a single Service
Provider to IPX connection
The model used for the Inter-Service Provider IP Backbone is that of a private IP backbone
network. All information is carried over these networks using the IP suite of protocols.
In this hierarchical model, Service Providers require only one connection and one agreement
with the Inter-Service Provider IP Backbone to be able to interconnect with selected Service
Provider partners. If redundancy is required, two or more physical connections to one or more
Inter-Service Provider IP Backbone networks may be used. (See Annex A for problems and
solutions to this approach). Service Providers obtain connections to Inter-Service Provider IP
Backbone nodes locally from an IP Backbone Provider or from other Providers (e.g. leased
lines).
A simplified high-level architecture of the GRX covering both roaming and interworking
interconnection cases is illustrated in the figure below.
Local
MNO A Tail
GRX GRX
Provider X Provider Y
Peering Interface
MNO C
Local
Tail
The GRX consists of separate and competing GRX Providers (or GRX Carriers). A GRX
network can be operated by any qualified party. Requirements for GRX Providers are described
in section 6.1.
GRX Providers connect to each other via peering interfaces. These peering interfaces may be
direct connections or may pass through a common peering point. GRX Providers should enter
into Service Level Agreements (SLAs) with other GRX Providers.
A common DNS root database supports domain name resolution. This root database may be
used by all GRX parties.
The GRX is isolated from the public Internet and security rules are defined to prevent
unintended access from it.
Dynamic routing may be used on the GRX using the BGP-4 routing protocol. When using the
GRX, interworking and/or roaming services can be established without specific configuration
requirements and additional processing of protocols.
It is possible to connect a Service Provider to the GRX using IPSec. This possibility is described
in the end-to-end QoS SLA. Other uses of IPSec are dependent on service requirements and
will be described in the service specific documentation, e.g. IR.61 for WLAN roaming.
The GRX offers a transport-only interconnection service between mobile operators on a bilateral
basis with no guarantees of QoS end-to-end. This transport-only function may be used to
transport any protocol on a bilateral basis. In particular, the GRX is used to support traffic
applications including: GPRS and 3G data roaming (using GTP), WLAN roaming authentication,
MMS Interworking and IMS Interworking. See section 9 for more information on traffic
applications.
The IPX builds upon and extends the architecture of the GRX by introducing a number of other
stakeholders – Fixed Network Operators, Internet Service Providers and Application Service
Providers (which together with Mobile Network Operators are hereafter termed Service
Providers). The IPX is formed from separate and competing IPX Providers. An IPX network can
be operated by any qualified party.
Local
Operator Tail
A
IPX IPX
Provider X Provider Y
Peering Interface
Operator
IPX IPX
Local
C
Proxy Proxy
Tail
End-to-end SLA
The IPX introduces the requirement to support Quality of Service features end-to-end. That is,
the parties involved in the transport of a service (up to the terminating Service Provider
BG/firewall) are bound by end-to-end Service Level Agreements.
The IPX also introduces IPX Proxy elements. These Proxies may support interworking of
specified IP services and make it possible to use cascading interconnect billing and a
multilateral interconnect model.
To assist with the translation of Telephone Numbers to URI the common DNS root database of
the IPX will support ENUM capability.
In the IPX, all user traffic, (that is, UE-to-UE and UE-to-Server), is separated from
Server-to-Server traffic. This is to fulfil the requirement of end users not being able to reach or
"explore" the IPX network.
Figure 3 above shows the high-level architecture of bilateral Service Provider traffic traversing
the Proxy element within Inter-Service Provider IP Backbone using any type of IP based traffic.
The user plane may or may not go through the proxy depending on each service requirement.
5.6 Types of Service Provider and Interconnectivity Allowed
There are three different types of Service Provider. They are classified according to the type(s)
of IP Backbone Provider(s) they connect to. This section describes each type and the
connectivity allowed between the different types.
5.6.1 MNO-G
This Service Provider is a GPRS/UMTS Mobile Network Operator who connects only to a GRX
network. The services they offer over the GRX network are on a bilateral basis with no
guarantees of QoS end-to-end. In this document they are called type MNO-G.
5.6.2 MNO-I
This Service Provider is a GPRS/UMTS Mobile Network Operator who connects to both a GRX
and an IPX network. They must be a GPRS/UMTS MNO because only these types of Service
Provider are allowed to connect to the GRX. In this document they are called type MNO-I.
5.6.3 NGNO
This Service Provider connects only to the IPX network and can be any type of organisation. In
practice they are unlikely to be a GPRS/UMTS Mobile Network Operator because GPRS/UMTS
roaming requires connection to a GRX Network in order to reach existing GPRS/UMTS Mobile
Network Operators. In this document they are called type NGNO.
Service Provider
MNO-G MNO-I NGNO
Type
9(GRX Services
MNO-G 9 8
Only)
9(GRX services
MNO-I 9 9 (IPX Services Only)
Only)
9 (IPX Services
NGNO 8 9
Only)
Table 1 – Interconnectivity Options
The GRX offers a transport-only interconnection service between Service Providers that are
GPRS/UMTS network operators on a bilateral basis with no guarantees of QoS end-to-end. This
transport-only function may be used to transport any protocol on a bilateral basis as well as
MMS hubbing and WLAN proxy services.
.
5.7.2 IPX Provider
An IPX Provider will be able to provide connections to any type of Service Provider – whether
GPRS/UMTS operator or not – and can carry both GRX and IPX services.
Connection to the GRX is restricted to GSM and 3GSM MNOs. In this section, the term Service
Provider above applies only to these parties.
GRX Providers should not act as a transit GRX. That is, GRX Providers should not pass traffic
over their network from one connected GRX network to another connected GRX network. A
packet should not pass through more than two GRX Providers’ networks.
All the requirements above relating to GRX Providers shall be supported by an IPX Provider. (In
this case, the term Service Provider in section 6.1 applies to MNOs, FNOs, ISPs and ASPs). In
addition, an IPX Provider shall:
• Support end-to-end QoS requirements, described in the end-to-end quality SLA and in
this document
• Create the agreements required with other IPX Providers to fulfill the end-to-end SLA
• Maintain user traffic separation as described in 6.5.6
IPX Providers shall not act as a transit IPX. That is, IPX Providers shall not pass traffic over
their network from one connected IPX network to another connected IPX network. A packet
shall not pass through more than two IPX Providers’ networks.
The interconnection allowed between IP Backbone Providers and Service Providers is shown in
the table below. Each connection must follow the rules shown. Note that a Service Provider may
connect to multiple IP Backbone Providers provided that each connection complies with the
rules in the table.
GRX 9 9 8
IPX 9 (GRX Services Only) 9 9 (IPX Services Only)
Table 2 - Interconnections between IP Backbone Provider and Service Provider
Inter-Operator
IP Backbone C
BGP Routing to
Layer 2
IP VPN
Operator and other
IP Backbone A- Inter-Operator IP
Backbones and
Inter-Operator IP Backbone C
DNS Root Service
IP Backbone A Frame Relay/ATM
PVCs
Layer 1
IP Backbone B –
IP Backbone C
Inter-Operator
IP Backbone B Layer 1 Link
Dynamic routing reduces the amount of management work in the event of a change IP address
requirements (i.e. new address ranges are applied). In addition, dynamic routing supports
redundant connections to IP Backbones/Service Providers.
IP Backbone Providers should exchange routing information and traffic between all other Inter-
Service Provider IP Backbone nodes. An IP Backbone Provider should be responsible for
distributing all Inter-Service Provider BGP-4 information to all its peers. An IP Backbone
Provider should advertise its customer networks to peering partners after a Service Provider has
fulfilled the security requirements laid down in PRD IR.77 [19]. When operating an IPX network,
the above requirements are mandatory.
The Service Provider and the IP Backbone Provider are both responsible for checking that all
connected Service Provider and IP Backbone Provider networks are invisible to and
inaccessible from the public Internet.
In an Inter-Service IP Backbone Provider environment with multiple peering points, it is
recommended that “Hot potato”[24] routing (where traffic is exchanged with the next IP
Backbone Provider at the nearest peering point in terms of delay) should be used. However
“Cold potato” routing (where traffic is exchanged with the next IP Backbone Provider at the
furthest peering point in terms of delay) may be agreed bilaterally between GRX/IPX Providers.
IP Backbone Providers shall not restrict protocols carried between Service Providers unless
those protocols are non-compliant with the requirements set out in the PRD IR.77 [19].
Inter-Operator
IP Backbone
BGP Extended community attributes are introduced as a tool to build this separation. This
section gives more precise descriptions of communities and how and where marking and
filtering is done in IPX networks. By giving each type of Service Provider its own dedicated BGP
extended community value we can ensure that, for example, a MNO-G Service Provider can
route traffic to another MNO-G Service Provider but not to an NGNO Service Provider. By using
extended community values coupled with a simple set of rules we can ensure that the IP
addresses of an MNO-G Service Provider are never advertised to an NGNO Service Provider
and vice versa. With other security rules like “no route of last resort” (see PRD IR.77[19]), this
also ensures that traffic from an NGNO Service Provider does not route via a GRX Provider.
This provides enhanced security for Service Providers and IP Backbone Providers.
Compliance with these rules is mandatory for IPX Providers. There is no mandatory requirement
for a GRX Provider to mark or check communities; these tasks are performed by IPX Providers.
GRX Providers may decide if they want to follow the guidance or not. GRX Providers may also
decide whether or not to keep these BGP communities untouched through their network.
Marking is always performed by the originating IP Backbone Provider, except in the case where
the originating IP backbone Provider is a GRX, when the marking in done by the first (and only)
IPX Provider in the path.
The following table lists the rules for marking of extended community values inside BGP4
routing between different types of Providers. It shows:
• the possible settings for the BGP4 extended community value marking
• whether marking the routing information is optional or mandatory
The regional info where the corresponding service providers are located and corresponding
values of nnnn are defined as follows:
The types of Service Provider and corresponding values of nnnn are defined as follows:
Among other security and routing definitions such as the prevention of transit Inter-Service
Provider IP backbone traffic, IPX Providers shall base their BGP-4 outbound advertisements for
BGP communities presented in section 6.5.3.
.
The following diagram demonstrates the full range of options for interconnection of Service
Providers and IP Backbone Providers. It does not demonstrate all possibilities for peering
between IP Backbone Providers.
In the GPRS roaming environment other communities can be in use and should be left
unhandled over the entire transport path. IP Backbone Provider should transparently transport
all communities, bilateral agreed between Service Provider and IP Backbone Provider, and
might add additional communities, but starting with their own AS number as described in section
6.5.3.2.
IPX D (NGNO)
MNO-G
NGNO only
MNO-I MNO-I
MNO-G NGNO
MNO-G
MNO-I
MNO-G
MNO-G
MNO-I,
MNO-I MNO-I
NGN
GRX
E (MNO-I)
GPRS and NGNO
MNO-G
MNO-I
MNO-G
Key:
MNO-I
Shows BGP Routing
G (MNO-G) Advertisement for
GPRS Only Type MNO-I IP
addresses
The following table (table 5) summarises the rules shown in the figure above. The table maps
destinations to communities, indicating which communities may be advertised by IPX Providers
to which destinations. Communities which are not permitted to be forwarded to a destination
shall be filtered out by the IPX Provider on that route.
6.5.5 DNS
As a minimum requirement, GRX Providers should support the transport of queries between
MNOs to allow for correct resolution of FQDNs for all service requirements, for example APNs
and MMSC hostnames (for MMS inter-working). IPX Providers shall support the transport of
such DNS queries. IPX Providers shall also provide for transport of ENUM queries to support
identified services.
The main specification for DNS of the Inter-Service Provider IP Backbone network is captured in
GSMA PRD IR.67 [17] with the exception of naming conventions which are specified in section
7.1.3.
Service Providers and IPX Providers shall also ensure that all UE IP datagrams are
encapsulated in tunnels to prevent the underlying IPX network from being reachable by end-
users.
UE-to-UE and UE-to-Server SIP/IMS IP traffic shall be encapsulated using GRE when
traversing the IPX. The encapsulation used for other types of UE IP datagrams shall be GTP for
GRPS roaming and IPSec for WLAN interworking. The encapsulation methods for other types of
UE IP datagrams are for further study.
Tunnels may terminate directly to other Service Providers, or may terminate at an IPX Proxy
(with a corresponding tunnel being used between the IPX Proxy and the terminating Service
Provider).
6.5.7 QoS
The GRX network may support Class of Service (CoS) parameters presented in section 8 of this
document.
The QoS requirements for the IPX are outlined in section 8 and also in the end-to-end QoS SLA
(IPX Agreement PRD AA.80 [22]). End-to-end QoS as described in [15] is a mandatory
requirement for IP Backbone Providers in the case of IPX.
Section 8 of this document concentrates on providing a traffic class specification and the
parameters for different classes of service.
The IPX will include a number of proxies that support specified IP service interworking. IPX
Proxies are not mandatory but will be needed to support Service Transit and Hubbing
Connectivity options. Note that the use of an IPX Proxy does not necessarily imply the adoption
of a multilateral connectivity model; Proxies may also be used to support services on a bilateral
basis. Specific requirements for each service will be captured elsewhere. The following is a non-
exhaustive list of generic features required from all IPX Proxies.
There are a number of practical advantages to using a Proxy. These include, but are not limited
to:
Service Providers may screen unwanted routes e.g. by selecting address ranges of their
interconnect partners.
7.1.2 Service Provider IP Addressing
Public addressing shall be applied in all Service Provider IP Backbone network elements, which
are advertised or visible to other Service Providers. Using public addressing means that each
Service Provider has a unique address space that is officially reserved from the Internet
addressing authority. However, public addressing does not mean that these addresses should
be visible to Internet. For security reasons, Service Provider and inter-Service Provider
backbone networks shall remain invisible and inaccessible to the public Internet.
Internet routers should not be able to route to the IP addresses advertised to the Inter-Service
Provider IP Backbone. In other words the IP Backbone Providers’ and Service Providers’
networks shall be totally separated from public Internet, from an IP routing perspective.
Currently, the Inter-Service Provider IP Backbone networks use IPv4 addressing and there is no
plan to introduce native IPv6 addressing in the foreseeable future. It is intended that IPv6 is
supported by tunnelling the IPv6 traffic over IPv4 between Service Providers where required.
Both IP Backbone Providers and Service Providers who employ IPv6 in their network should
assume full responsibility for all network adjustments necessary for maintaining connectivity to
all other IP Backbone Providers and/or Service Providers that deploy IPv4.
7.1.3 Service Provider DNS
The main specification for DNS of the Inter-Service Provider IP Backbone network is captured in
GSMA PRD IR.67 [17] with the exception of naming conventions which are specified below.
Naming Conventions
Having a consistent naming convention makes it easier for tracing and trouble shooting as well
as easing the maintenance of the DNS (see GSMA PRD IR.67 for more information on DNS on
the Inter-Service Provider IP Backbone). The following convention is recommended to achieve
these goals. Although the usage of this naming methodology is highly recommended, it is not
mandated.
All applicable Inter-Service Provider IP Backbone components that require address resolution
should be included in DNS. See GSMA PRD IR.67 [17] for more information.
Service Provider nodes should have names for each interface with the following format:
<city>-<type>-<nbr>[-<interface-type>-<interface-code>]
(the square brackets denote an optional part, which can be useful for router interface
information):
where:
• <city> is the name, or shortened name, of the city/town (or closest, where applicable)
where the node is located
• <nbr> is a running number of similar devices at the same city (for DNS servers, use 0 to
indicate the primary DNS Server)
• <type> describes device type and should be one of the following in the case of GRX:
• dns
• ggsn
• sgsn
• rtr - router
• fw - firewall
• <interface-type> has a couple of characters describing the interface type
• e - ethernet
• fe - fast ethernet
• ge - gigabit ethernet
• t - token ring
• s - serial
• h - HSSI
• a - ATM
• <interface-code> identifies interface slot/card/port etc.
Additional values for the <type> parameter are for further study for the IPX.
For example, the following are valid hostnames for interfaces on Service Provider nodes:
• helsinki-ggsn-4
• helsinki-rtr-2-fe-0-1
The domain name to append to hostnames for nodes belonging to MNO Service Providers
should be one of the following (where <MNC> and <MCC> are the MNC/MCC assigned to that
Service Provider):
• mnc<MNC>.mcc<MCC>.gprs
• mnc<MNC>.mcc<MCC>.3gppnetwork.org
A combination could be used by a Service Provider; however, for consistency it is better to use
only one. For more information on domain names, including the exact format and encoding, see
GSMA PRD IR.67 [17].
Access Point Names
Access Point Names (APNs) are defined in 3GPP TS 23.003 [18], Section 9. Access Point
Names are not case sensitive. The name consists of two parts
Network ID Operator ID
The Operator ID consists of MNC and MCC codes derived from the serving subscriber’s IMSI,
and uses the ".gprs" top level domain. Optionally, a Service Provider may provide a more
human readable Operator ID format, for example "sonera.fi". For more information on the
Operator ID part of the APN and rules of usage, see GSMA PRD IR.67 [17].
The Network ID consists of a label indicating to which network the subscriber is to be connected
e.g. the Internet, private office LAN, and so on. The Network ID should be either a registered
organization domain name (e.g. "example.com", "example.fi", "example.co.uk" and so on) or
Service Access Point Name such as "internet".
It should be noted that the MS never gives the Operator ID; this is appended to the Network ID
received from the MS by the SGSN. The MS may also not provide a Network ID, in which case
a default Network ID is determined by the SGSN, before appending the Operator ID. For more
information on APN selection rules, see 3GPP TS 23.060 [3], Annex A.
For example, the following are valid APNs that could be sent out by an SGSN to DNS:
• example.com.mnc091.mcc244.gprs
• example.com.sonera.fi
• internet.mnc091.mcc244.gprs
• internet.sonera.fi
7.1.4 Service Provider Security and Screening
Service Providers shall meet the requirements for security laid out in PRD IR.77.
It is strongly recommended that Service Providers implement firewalls at the ingress points of
their networks; for mobile operators, that is adjacent to Border Gateways. It is further
recommended that Service Providers using the IPX implement a firewall function to prevent
packets with incorrect/invalid IP addresses from being passed onto the IPX.
Each Service Provider shall be responsible for screening traffic inbound to its own BG/ Firewall.
Generally, Service Providers should allow only protocols that are needed for established
services, troubleshooting and network monitoring. Note that 'ping' and 'traceroute' are mainly
used for testing, troubleshooting and QoS measurement purposes. The end-to-end QoS SLA
(QoSi SLA 15) describes different options for measurements over an IPX backbone, including
local tails used by both Service Providers. A description and a usage policy for diagnostic tools
should be included in the interconnect agreement.
Service Providers shall ensure that for all IPX connections, all user traffic, (UE-to-UE and
UE-to-Server), is separated from Server-to-Server traffic. This is to fulfil the requirement of end
users not being able to reach or "explore" the IPX network.
Service Providers shall also ensure that all UE IP datagrams sent to the IPX are encapsulated
in tunnels to prevent the underlying IPX network from being reachable by end-users. Tunnels
may terminate directly to other Service Providers, or may terminate at an IPX Proxy (with a
corresponding tunnel being used between the IPX Proxy and the terminating Service Provider).
UE-to-UE and UE-to-Server IP traffic shall be encapsulated using GRE when traversing the IPX
except for GPRS Roaming (where GTP is used) and WLAN interworking (where IPSec is used).
7.2 BGP Advertisement Rules
7.2.1 General Rules
Service Provider’s core network addresses may be advertised to the Inter-Service Provider IP
Backbone with the BGP-4 [6] routing protocol and shall have an AS (Autonomous System) [6]
number acquired from the Internet addressing authority or the GSMA as appropriate. The
acquired AS number should be used as an originating AS when a Service Provider advertises
IR.34 Version 4.4 Page 27 of 52
GSM Association
Official Document: IR.34 Unrestricted
its own IP addresses to the Inter-Service Provider IP Backbone. When connecting to an IPX, the
BGP-4 protocol shall be used for advertising a Service Provider’s network addresses.
Service Providers using the GRX should follow the BGP advertisement style rules listed below.
• No host specific route or small block advertisements shall be advertised to the Inter-
Service Provider IP Backbone. No mask/29 advertisements shall be made unless a
smaller block was allocated to that Service Provider by the internet registry
• Advertised routes from each Service Provider shall be summarized whenever possible.
Summarizing smaller subnets into bigger blocks will minimize size of the routing tables
and reduce router processing load. This summary may be carried out by the Service
Provider or the IP Backbone Provider
• Service Providers shall only advertise their own core public IP address ranges into the
Inter-Service Provider IP Backbone
• Networks advertised by Service Providers shall originate from the AS number assigned
to them. (AS path shall start Service Provider AS number)
• Service Providers must only notify to their IP Backbone Providers(s) IP address ranges
used by their network infrastructure. This allows their Providers to build their routing
filters
• Service Providers shall use BGPv4 communities presented in chapter 6.5.3, to tag
all it’s own network advertisements towards Inter-Service Provider IP Backbone
•
• IP address ranges used by User Terminals must not be advertised to or routed on the
Inter-Service Provider IP Backbone
Service Providers using the IPX shall follow the rules above.
The BGP advertisements of Services Providers will be marked by IPX Providers according to
the rules described in section 6.5.3. No action is required from the Service Provider.
The end-to-end SLA [22, Annex] describes the different options for establishing physical
connections from a Service Provider to the IPX. Different connection options can be divided into
three categories:
• Layer 1 connection (e.g. leased line or fibre) or
• Layer 2 logical connection (e. g. ATM, LAN, Frame Relay) or
• Layer 3 IP VPN connection over public IP network (IPSec is recommended)
The use by a Service Provider of an Internet IPSec VPN for the local tail is strongly discouraged
unless there is no viable alternative.
It is up to IP Backbone Provider and Service Provider to determine the exact details of each
connection bilaterally, however details in the end-to-end SLA shall be followed in the case of
IPX. The main benefits of the Inter-Service Provider IP Backbone structure to Service Providers
are:
• The Service Providers do not have to create dedicated connections to every roaming
partner. One connection to one Inter Service Provider IP Backbone Provider is required
as a minimum.
IR.34 Version 4.4 Page 28 of 52
GSM Association
Official Document: IR.34 Unrestricted
• Service Providers may choose to start with low quality and low capacity connection to
the Inter-Service Provider IP Backbone and upgrade the level of connectivity when it is
economically feasible and there are traffic volumes and type of traffic that require more
bandwidth and better quality i.e. the IP Backbone is scalable and able to meet Service
Provider Requirements.
• Inter-Service Provider IP Backbone has QoS implemented and shall be measurable on a
Service Provider by Service Provider basis.
• Inter-Service Provider IP Backbone introduces a Hub Connectivity Option to simplify
different interworking scenarios
Operator A Layer 2
network Operator B
Border Frame Relay/ATM
GW PVCs Layer 1
Operator B Operator C
network Border
GW Layer 1 Link
Operator C
network
8 QOS
8.1.2 Responsibilities
Terms and conditions of each SLA component should be examined and whether Service
Provider's account should be credited and if so to extent where the SLA has not been met. Note
penalties shall be in scope for any future governance.
Help desk support and customer services should be considered.
It is recommended that following traffic classes are available and marked as presented in the
table 7. Traffic classes are distinguished by Differential Service bits. These bits are seen in IP
ToS field and used for prioritizing traffic if needed.
QoS Information
Diffser
DSCP
v PHB
Traffic Class THP
Conversationa
N/A EF 101110
l
1 AF31 011010
3 AF11 001010
Service providers are responsible for marking packets to correct traffic classes. They may
outsource this functionality to Inter-Service IP Backbone provider when suitable. Inter-Service IP
Backbone providers may change DSCP values in their own network as long as they return
values set by operators before traffic is given to an other inter-Service IP Backbone provider or
Service Provider and they fulfil given values for parameters per class
8.3.2 Delay
Roundtrip delay is the total time that it takes to transmit an IP packet from the SOURCE to the
Destination and receive the reply packet from the destination at the SOURCE. (Measured over
a given period of time, in milliseconds)
Table 8 and 9 presents Roundtrip delay values between originating and terminating Service
Provider premises.
It should be noted that actual performance of Inter Service provider IP backbone network could
be better than given reference values in the table 8 an 9.
South- N America
Middle- North- East South East N America Central
East Oceania (West S America Africa
Europe Europe Europe Europe Asia (East Coast) America
EF & AF4 Asia Coast)
Middle-
Europe 55 45 80 72 340 360 380 120 200 225 330 242
East Europe 80 35 40 102 360 370 420 165 215 281 350 262
South Europe 72 75 102 72 345 355 380 145 220 247 335 218
East Asia 340 350 360 345 150 165 275 340 285 353 460 383
South-East
Asia 360 360 370 355 165 145 255 360 310 489 480 251
Oceania 380 400 420 380 275 255 90 360 310 369 470 287
N America
(East Coast) 120 130 165 145 340 360 360 40 90 92 280 326
N America
(West Coast) 200 215 215 220 285 310 310 90 40 126 300 418
Central
America 225 249 281 247 353 489 369 92 126 40 137 294
S America 330 335 350 335 460 480 470 280 300 137 120 180
Africa 242 269 262 218 383 251 287 326 418 294 180 180
Table 8. Delay values for conversational and streaming traffic classes
South- N America
Middle- North- East South East N America Central
East Oceania (West S America Africa
Europe Europe Europe Europe Asia (East Coast) America
AF1-3 + BE Asia Coast)
Middle-
Europe 66 54 96 86 408 432 456 144 240 270 396 290
East Europe 104 42 48 122 432 444 504 198 258 337 420 315
South Europe 94 90 122 86 414 426 456 174 264 297 402 262
East Asia 442 420 432 414 180 198 330 408 342 424 552 459
South-East
Asia 468 432 444 426 198 174 306 432 372 587 576 301
Oceania 494 480 504 456 330 306 108 432 372 442 564 345
N America
(East Coast) 156 156 198 174 408 432 432 48 108 111 336 391
N America
(West Coast) 260 258 258 264 342 372 372 108 48 151 360 501
Central
America 292 298 337 297 424 587 442 111 151 48 165 352
S America 429 402 420 402 552 576 564 336 360 165 144 216
Africa 314 322 315 262 459 301 345 391 501 352 216 216
8.3.3 Jitter
Jitter (or the IP Packet Delay Variation as it may be known) is the delay variation among the
different packets sent from the SOURCE to the DESTINATION (Measured over a given period
of time, in milliseconds.) and measured as follows.
Definition
Standards-based definition: IETF RFC 3393, IP Packet Delay Variation Metric (IPPM);
RFC 3393 states, “A definition of the IP Packet Delay Variation (ipdv) can be given for packets
inside a stream of packets.
The ipdv of a pair of packets within a stream of packets is defined for a selected pair of packets
in the stream going from one measurement point MP1 to another Measurement point MP2. In
this case the measurement points are the same as those that have been defined for Delay, as
outlined in section 8.3.2.
The ipdv is the difference between the one-way-delay of the selected packets.”
RFC 3393 states that measuring jitter from a source to a destination host is useful for the
following reasons:-
• One important use of delay variation is the sizing of play-out buffers for applications
requiring the regular delivery of packets (for example, voice or video play-out). What is
normally important in this case is the maximum delay variation, which is used to size
play-out buffers for such applications;
• Other uses of a delay variation metric are, for example, to determine the dynamics of
queues within a network (or router) where the changes in delay variation can be linked to
changes in the queue length process at a given link or a combination of links;
• In addition, this type of metric is particularly robust with respect to differences and
variations of the clocks of the two hosts. This allows the use of the metric even if the two
hosts that support the measurement points are not synchronized.
The following Jitter values shall only apply to conversational and streaming traffic classes (i.e.
EF and AF4 traffic classes).
Intra-continent Jitter Value - 5mS per GRX/IPX Provider (maximum of 2 involved in the service
delivery chain)
Inter-continent Jitter Value - 10mS per GRX/IPX Provider (maximum of 2 involved in the service
delivery chain)
Intra-Continent Traffic
In the case where traffic is exchanged over one GRX/IPX Provider between Service Providers in
the same Continent, the total end to end Jitter value would be 5mS. This would increase to
10mS (5mS x 2) where 2 GRX/IPX's are involved in the service delivery chain in that Continent).
Inter-Continent Traffic
In the case where traffic exchanged between Service Providers in different continents, and
GRX/IPX1 is exchanging traffic with GRX/IPX2 in the same continent as the originating Service
Provider, GRX/IPX 1 would have a 5mS Jitter target and GRX/IPX2 would have a 10mS target
to recognise that IPX/GRX2 traffic is inter-continental.
AF1 <0.1%
EF + AF4 0.1%
9 TRAFFIC APPLICATIONS
The following sections describe some of the traffic applications for the Inter-Service Provider IP
Backbone network.
GGSN SGSN
SGSN GGSN
Inter-Operator
IP Backbone
9.5 IMS
The Inter-Service Provider IP Backbone can be used for IMS interworking [20] between IMS
networks [21] as depicted in Figure 10 below. Note that User Plane traffic may or may not be
sent through the IPX Proxy.
IMS interworking will introduce new protocols (e.g used by peer-to-peer applications in the user
plane) which the IP Backbone Provider shall not restrict. User-to-User or User-to-Server traffic
shall be carried inside GRE tunnel over the Inter-Service Provider IP Backbone. At least the
User Plane shall be encapsulated, and it is optional whether or not the Control Plane is
encapsulated. For further information on IMS see PRD IR.65 IMS Roaming & Interworking
Guidelines.
More detailed requirements for an IPX Proxy for SIP-based traffic can be found in Annex B:
Proxy requirements.
10 REFERENCES
[1] GSMA PRD IR.33: "GPRS Guidelines"
[2] GSMA PRD IR.35: "End to End Functional Capability specification for Inter-
Operator GPRS Roaming"
[3] 3GPP TS 23.060: " General Packet Radio Service (GPRS); Service Description;
Stage 2"
[4] 3GPP TS 21.905: "3G Vocabulary"
[5] IETF RFC 1812: "Requirements for IP Version 4 Routers"
[6] IETF RFC 4271: "A Border Gateway Protocol 4 (BGP-4)"
[7] 3GPP TS 29.060: " General Packet Radio Service (GPRS); GPRS Tunnelling
Protocol (GTP) across the Gn and Gp Interface"
[8] IETF RFC 4301: "Security Architecture for the Internet Protocol"
[9] IETF RFC 4302: "IP Authentication Header"
[10] IETF RFC 4305: Cryptographic Algorithm Implementation Requirements for
Encapsulating Security Payload (ESP) and Authentication Header (AH)
[11] IETF RFC 4303: "IP Encapsulating Security Payload (ESP)":
[12] IETF RFC 4306: "The Internet Key Exchange (IKE)"
[13] 3GPP TS 23.107: “3rd Generation Partnership Project; Technical Specification
Group Services and System Aspects; QoS Concept and Architecture”
[14] http://www.ietf.org/html.charters/diffserv-
charter.htmlhttp://www.ietf.org/html.charters/OLD/diffserv-charter.html
[15] IETF RFC 3246: "An Expedited Forwarding PHP"
[16] IETF RFC 2597: "Assured Forwarding PHB Group"
[17] GSMA PRD IR.67: "DNS Guidelines for Operators"
[18] 3GPP TS 23.003: "Numbering, addressing and identification"
[19] GSMA PRD IR.77 “Inter-Operator IP Backbone Security requirements For
Service Providers and Inter-Operator IP Backbone Providers”
[20] GSMA PRD IR.65 “IMS Roaming & Interworking Guidelines”
[21] 3GPP TS 23.228: “IP Multimedia Subsystem (IMS); Stage 2”
[22] GSMA PRD AA.80 “General Terms & Conditions For Agreement for IP Packet
eXchange (IPX) Services”
[23] IETF RFC 4360: “BGP Extended Communities Attribute”
[24] IETF RFC 4277: “MEDs and Potatoes”
Inter-Operator
IP Backbone
BG/FW
BG/FW
Operator A Inter-Operator
Operator B
IP Backbone
BG/FW
Inter-Operator
IP Backbone
It is recommended that Service Providers agree with their own Inter-Service Provider IP
Backbone Providers how backbone addresses are advertised to the IP Backbone Provider(s) of
the participating Service Provider.
As shown in the figure asymmetrical routing causes FW (IP security device) problem on the
Service Provider side since firewall state information of BG1 is typically not available on BG2.
The packets will be dropped. Thus, the network design of the Service Provider is the source of
the problem.Therefore, the Service Provider itself should implement such a network design
within its network, which can avoid the "double Inter-Service Provider IP Backbone problem".
If the "double Inter-Service Provider IP Backbone problem" applies, Service Providers have two
options:
Inter-Operator
IP Backbone
BG/FW
BG/FW
Operator A Inter-Operator
Operator B
IP Backbone
BG/FW
Inter-Operator
IP Backbone
Figure 13 - Proposal for network design to overcome the "double Inter-Service Provider
IP Backbone problem"
Since the FWs are located behind the both BGs the "double Inter-Service Provider IP Backbone
problem" is solved. This network design allows unlimited future scalability if the network grows.
The following figure shows a possible future network design. It shows an Service Provider
network with different sites. Every site has its own IP range, which is routed in the backbone.
The Inter-Service Provider IP Backbone has QoS, the Service Providers need to define precise
routing policy between Service Provider networks to account for this requirement
Inter-Operator
FW
IP Backbone ‘1’
active
Site ‘A’
BG ‘1’
FW
active
Inter-Operator
BG ‘2’
IP Backbone ‘2’ Site ‘B’
Gp
Backbone
Operator
BG ‘n’
FW
active
Inter-Operator IP
Backbone ‘1’ BG/FW
IP – Range 1
IP – Range 2
Inter-Operator IP BG/FW
Backbone ‘2’
Operator
GTP connection it does not have to attempt using GTPv1 if it already knows that the destination
does not support it.
It is therefore recommended that Operators make full use of such tables within SGSNs. Doing
this will reduce the number of re-attempts that have to be made to establish a GTP connection.
Home Operator:
Many firewalls are configured to simply "drop" packets (i.e. do not send back any error to the
sender) destined for ports which do not have a service running on them. This means that a
GTPv1 capable SGSN in a foreign network trying to contact a GTPv0 only GGSN in a
subscriber's home network will have to wait for a specific period of time before re-attempting the
connection using GTPv0. The same applies for Inter-MNO Operator IP handover when the
SGSN in the old network supports GTPv1 and the SGSN in the new network supports only
GTPv0.
It is therefore recommended that Operators who do not yet support GTPv1 configure their
firewalls on their GGSNs (and/or any border gateways at the edge of the network) to "deny"
packets destined for the GTPv1 signalling/control plane port (UDP/IP port 2123) by sending
back ICMP message 3 "destination unreachable" with error code 3, "Port unreachable". Doing
this will greatly reduce the time taken for an SGSN to realise that the destination does not
support GTPv1.
Therefore, at this time, the service level of the Inter-Service Provider IP Backbone will be
defined by IP service QoS parameters described below. Service Providers and IP Backbone
Providers are encouraged to monitor the development of improved IP QoS technologies, such
as ‘Differentiated Services (diffserv)’ model [14].
Mapping of the GPRS Release 97 and Release 99 QoS classes into IP service QoS parameters
will be necessary later. Forthcoming GPRS release specific QoS issues should remain open for
further study.
For data roaming taking place between two networks of different generations, i.e. 3G (GPRS
R99/UMTS) and 2.5G (GPRS R97/98), Service Providers should comply with the IP QoS
definitions for GPRS R97/98.
B1 INTRODUCTION
For those Service Providers wishing to make use of the “hub” approach in IPX, an IPX Proxy will
be provided. An IPX Proxy is located within the IPX routing domain. An IPX Proxy has features
and functions from traditional SIP proxy as well from B2BUA equipment to meet the Service
Provider requirements, as described in this document. However, IPX Proxy is neither a pure
B2BUA nor SIP proxy node as such.
The aim of this Annex is to list the technical requirements for IPX Proxy, regarding issues such
as support for different IP versions, control plane support, user plane support, transcoding
capability, routing ability and security related functions.
Note that in order to achieve maximum feasibility of this annex (as opposed to purely high-level
theoretical wishes), these requirements are aimed to be as simple as possible. This also helps
to maintain future-proofing of IPX Proxy.
Only the functional parts of IPX Proxy belong to this particular annex, physical aspects of the
IPX Proxy implementation are out of scope.
In this Annex IPX Proxy is just a name for a certain kind of functionality and does not directly
reference to RFC 3261 or any other RFC’s.
R1. IPX Proxy shall be able to act as transparently as possible towards connected parties,
either as a transparent proxy or a known proxy. No modifications to standard
interworking/interconnection interfaces need to be done because of IPX Proxy
I.e. the IMS core system does not need to be altered due to the introduction of an IPX Proxy.
Also from an end-user point of view the inclusion of IPX Proxy shall be transparent. If there is a
need to perform transcoding or other kind of conversion regarding a part of IP packet, other
parts of that packet remain unchanged. More detailed requirements related to this subject are
listed later in sections B2.1 – B2
R2. IPX Proxy shall be able to handle inter-Service Provider traffic in a secured and controlled
manner
This means that the overall IPX Proxy infrastructure is protected from outsiders. Requirements
for the IPX network are listed in section 6.5.
R3. Common, agreed interfaces shall be used in all IPX Proxy connections
R5. IPX Proxy shall be able to act in a service agnostic manner, when possible
Note that in certain cases, e.g. when transcoding is performed by an IPX Proxy, this
requirement does not apply. It shall not be necessary to modify settings in an IPX Proxy due to
new service using standard protocols being introduced in inter-service Service Provider
interface using IPX. The basic design principle of IPX Proxy is that it supports all kinds of
various user plane protocols, i.e. if new service works over direct IMS-to-IMS connection, it shall
work also when traffic is routed via IPX Proxy
R6. IPX Proxy shall be able to relay both user plane and control plane. IPX Proxy shall also be
able to relay only control plane if so desired, based on contractual arrangement
I.e. user plane can be separated from control plane and routed directly between originating and
terminating Service Providers in some cases
R7. IPX Proxy shall be able to relay traffic between terminals & servers using different
addressing schemes (such as between IPv6 and IPv4 addresses)
I.e. IPX Proxy shall have dynamic port & address translation capabilities
R8. IPX Proxy shall be able to verify that source is who it pretends to be
E.g. verifying CSCF IP addresses with the IP addresses stored in the local IPX Proxy database.
Source refers to the originating Service Provider, not the end-user. Traffic should be dropped in
case source cannot be verified
R9. IPX Proxy shall know capabilities of terminating Service Provider and ensure media is
appropriately handled
R10. IPX Proxy shall be able to be used as ”masking device” by Service Providers (such as
mobile Operators) towards other Service Providers
I.e. Service Provider A should not need to modify its own configuration due to connection with
Service Provider B, C or D, regardless of what kind of IP versions, IP address ranges, routing
domains or protocols B, C and D are using. IPX Proxy can be used as a single point of contact
for inter-Service Provider traffic, regardless of the number of interworking partners
R11. IPX Proxy may be able to verify that next hop is reachable (e.g. on IP layer)
R12. IPX Proxy shall have reporting capabilities, regarding e.g. Proxy performance
I.e. it shall be possible to configure IPX Proxy in such a way that it is possible for control plane &
user plane traffic being transmitted from Service Provider A’s UE1 to Service Provider A’s UE2
via IPX Proxy. In practice this involves IPX Proxy DNS tables. With loopback testing it is
possible for Service Provider A to test IPX Proxy without any other parties involved
R15. IPX Proxy shall have DNS resolver capability for ENUM functionality (and generally for
NAPTR and SRV RRs)
R16. IPX Proxy shall be able to support DiffServ packet queuing
R17. IPX Proxy shall have enough processing power to minimize delays, also in case that user
plane is routed through IPX Proxy
I.e. IPX Proxy shall not be a bottleneck in inter-Service Provider connection, overall session
performance shall not significantly degrade due to inclusion of IPX Proxy
R18. IPX Proxy shall have external interface(s) towards external management system and O&M
system
R20. IPX Proxy shall be able to support external interface(s) towards billing system
E.g. in the case of PoC interconnection, IPX proxy shall be capable of translating between
3GPP AMR and 3GPP2 EVRC codecs, if no other network node (such as PoC server or IMS
core system) or terminal is capable of providing this functionality. Unnecessary transcoding shall
be avoided, i.e. transcoding shall not be “automatically” applied in cases when it is not required
(e.g. when operating within purely 3GPP compliant environment or transcoding is performed by
other elements)
R22. IPX Proxy shall know destination Service Provider capabilities on control plane & user
plane level, for transcoding/conversion purposes
SIP proxy mode can apply if only the control plane is routed via IPX Proxy
IR.34 Version 4.2 Page 48 of 52
GSM Association
Official Document: IR.34 Unrestricted
R24. IPX Proxy shall be able to connect to various different servers and core systems, such as
IETF SIP servers and 3GPP IMS core systems
R25. IPX Proxy shall be able to support 3GPP standards compliant interfaces for IMS
connectivity (Mw for control plane & Gi/Mb for user plane)
R26. IPX Proxy shall support SIP error codes as specified by IETF & 3GPP
R27. It shall be possible to configure IPX Proxy in such a way that it always forwards any
unknown SIP methods/headers/parameters towards recipient unmodified
This can be used e.g. to support new SIP extensions. However, IPX Proxy should make a note
of this activity into the log file, in case this is used for malicious purposes
R28. IPX Proxy shall be able to modify IP addresses in SIP/SDP messages when it is acting as
a media proxy
R29. IPX Proxy shall be able to modify SIP headers (fields such as Via, Contact, Record Route,
Content-Length) when it is acting as a media proxy
R30. IPX Proxy shall be able to support conversions between different common protocols and
protocol versions, such as translate between wireline and wireless SIP variants
It is strongly recommended to utilize only standardized SIP profiles, but it is seen that in the
market also some non-standardized SIP profiles & extensions will be anyhow used. Therefore
IPX Proxy implementations should take this into account somehow. Note that conversion shall
not be “automatically” applied in cases when it is not required (e.g. when operating within purely
3GPP compliant environment or conversion is performed by other elements)
R32. IPX Proxy shall be able to relay traffic between terminals that are located in different
networks and use overlapping private IPV4 addresses
Private IPV4 based addressing is widely used. It cannot be assumed that mobile Operators co-
ordinate their private IPV4 address ranges, therefore IPX Proxy has to be able to handle it
R33. Only IP addresses that are routable in IPX shall be used in inter-proxy interface (for any
IPV4 based GRE tunnel outside address)
Interfaces from and to IPX Proxy (including Proxy-to-Proxy and Service Provider-to-Proxy
connections) shall not use private IP addresses or public IP addresses routable in internet
R34. IPX Proxy shall be able to support IPV6 connectivity without any IP version related
modifications, i.e. IPV6 based Service Provider shall be able to connect to another IPV6 based
Service Provider without any kinds of IPV6-to-IPV4 conversions done in between
I.e. end-to-end IPV6 connections shall be possible
R35. IPX Proxy-to-IPX Proxy connection shall be capable of using IPV6 if proxies are routing
traffic between two IPV6 based Service Providers or between IPV6 based Service Provider and
IPV4 based Service Provider
R36. IPX Proxy-to-IPX Proxy connection shall be capable of using IPV4, only if proxies are
routing traffic between two IPV4 based Service Providers
R37. Proxy shall not modify IPV6 based IP addresses in the user plane (if no IPV4 related
conversion is needed)
R38. IPX Proxy shall be able to support tunnelled traffic (e.g. GRE) and non-tunnelled traffic, for
both control and user planes, including inter-IPX Proxy interface
Control plane & user plane can be in the same tunnel or use separate tunnels. Control plane &
user plane can be also un-tunnelled
R39. IPX Proxy shall be able to store routing information, regarding IP address/port pair used to
receive a particular media stream and the destination address/port pair necessary to forward the
media to its ultimate destination
R40. IPX Proxy shall support UDP and TCP as transport protocol, along with unlimited number
of possible media/application protocols (such as RTP, RTCP, HTTP, MSRP etc)
In other words, IPX Proxy should not place any restrictions to supported protocols. Also the port
numbers used by IPX Proxy shall be configurable
R41. IPX Proxy shall be able to forward media streams and perform termination & initiation of
media streams, when functioning as a B2BUA
R42. IPX Proxy shall support both SIP URI and tel URI end-user addressing schemes
IPX Proxy shall be capable of making ENUM queries, if not handled by other network node. IPX
Proxy shall also be able to pass incoming tel URI unmodified through, if needed.
R44. Maximum delay for signalling caused by IPX Proxy shall be less than 20ms per SIP
message
R45. Maximum delay for media caused by IPX Proxy shall be less than 1ms (when transcoding
is not applied)
R46. IPX Proxy should introduce no more jitter than the latency figure for media flows (see R45.
for maximum delay)
R48. IPX Proxy must be able to relay TOS (Type of Service) field of the IP header from source
to destination unmodified
It shall be also possible to modify TOS bits by IPX Proxy, since in some cases e.g. source and
destination use different style of TOS mapping and IPX Proxy needs to make sure that mapping
is correctly used
R53. IPX Proxy shall be able to perform secure NAT traversal as well as firewall traversal for
signalling and media, when needed
R54. IPX Proxy shall support opening pinholes for user plane traffic traversal based on SIP/SDP
information
R55. IPX Proxy shall support closing pinholes used by user plane traffic based on SIP/SDP
information
It shall be also possible to close these pinholes automatically e.g. when SIP Session Timer
expires (for example in case of connectivity problems)
R56. IPX Proxy shall block traffic not related to ongoing SIP sessions
I.e. SIP session set-up needs to be done before media is allowed
R57. IPX Proxy shall be able to have rate limit / flow control features on control plane as well as
on user plane
Including capability for propagation prevention, e.g. prevent problems such as flooding from one
network to other connected networks. Flow control can be applied both to control and user
plane, on a per Service Provider basis. It shall be possible to configure alarms for the amount of
traffic, in order to prevent overloading
R58. IPX Proxy shall support the ability to apply admission control per domain basis
R59. IPX Proxy may support the ability to support maximum admission control limits per domain
basis
Helps e.g. preventing DoS attacks by setting a maximum limit of simultaneous connections
R60. IPX Proxy shall be able to handle policy function across domains
For example bandwidth control & admission control
R61. IPX Proxy shall be able to support user plane policing based on the data rate
R62. IPX Proxy should be able to support external interface(s) towards policy control/admission
control function
R63. IPX Proxy shall be able to check that media is what session setup implies
For fraud prevention purposed it is important to be able to check that e.g. session set up as PoC
really contains PoC related media instead of something else
R65. IPX Proxy shall be able to produce Inter-Service Provider charging data based on at least
control plane (SIP / SDP)
I.e. either control plane or user plane or both can be used as a basis for Inter-Service Provider
charging data
R66. IPX Proxy shall be able to produce service related charging data
E.g. IPX Proxy shall be able to generate Inter-Service Provider charging data for OMA PoC
based on RTCP messages used in POC-4 interface
R67. IPX Proxy should provide a means to correlate the charging information generated at
Transport, Service and Content levels by various entities
R68. IPX Proxy should support different Charging objects for generating charging data. It should
be configurable to define these objects and their properties/attributes