Considerations with WebRTC in
Mobile NetworksCisco Systems, Inc.Cessna Business Park, Varthur HobliSarjapur Marathalli Outer Ring RoadBangaloreKarnataka560103Indiatireddy@cisco.comHuawei5340 Legacy Drive, Suite 175Plano Texas 75024john.kaippallimalil@huawei.comCisco Systems, Inc.Cessna Business Park, Varthur HobliSarjapur Marathalli Outer Ring RoadBangaloreKarnataka560103Indiarmohanr@cisco.comAlcatel-Lucent1960 Lucent LaneNaperville, Illinois 60563-1594USrichard.ejzak@alcatel-lucent.comRTCWEBThis document discusses some of the issues with WebRTC applications
in Mobile Networks.The use of cellular broadband for accessing the Internet and other
data services via smartphones, tablets, and notebook/netbook computers
has increased rapidly as a result of high-speed packet data networks
such as HSPA and HSPA+; and now Long-Term Evolution (LTE) is being
deployed. Browsers on these devices are becoming similar in capability
to their desktop counterparts. So, from that perspective, it is feasible
to run WebRTC applications in them. This draft enumerates concerns when
real time applications like WebRTC are used on such devices in the above
listed networks.This note focuses on QOS and traffic offload aspects, and does not
address other mobile network related topics like power consumption,
interface switching and congestion control related issues already being
discussed in .The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in .This note uses terminology defined in UE, MS, MN, and Mobile : The terms UE (User Equipment), MS (Mobile
Station), MN (Mobile Node), and mobile refer to the devices that are
hosts with the ability to obtain Internet connectivity via a 3GPP
network.WebRTC Server : Web Server which supports WebRTC.This document can be used as a tool to design solution(s) for WebRTC
QoS and traffic offload in cellular broadband networks. It describes key
use cases, factors common to the use cases, and key solution elements.
This note aids WebRTC server designers and network architects to
facilitate proper deployment of WebRTC in mobile networks. The WebRTC
server could either be deployed in the Mobile Network, in a 3rd party
network trusted by the Mobile Network, or in a public network as a
Public WebRTC server.3GPP has standardized QoS for EPC (Enhanced Packet Core) from Release
8 . 3GPP QoS policy configuration defines
access agnostic QoS parameters that can be used to provide service
differentiation in multi vendor and operator deployments. The concept of
a bearer is used as the basic construct for which QoS treatment is
applied for uplink and downlink packet flows between the MN and gateway
. A bearer may have more than one packet
filter associated and this is called a Traffic Flow Template (TFT). IP
source address, source port, IP destination, destination port, L4
protocol, Type of service/Traffic class type, Security parameter index
etc identify a packet filter. Each MN can have one or multiple bearers
associated with its registration, each supporting different QoS
characteristics. An UpLink Traffic Flow Template (UL TFT) is the set of
uplink packet filters in a TFT. A DownLink Traffic Flow Template (DL
TFT) is the set of downlink packet filters in a TFT.The access agnostic QoS parameters associated with each bearer are
QCI (QoS Class Identifier), ARP (Allocation and Retention Priority), MBR
(Maximum Bit Rate) and optionally GBR (Guaranteed Bit Rate). QCI is a
scalar that defines packet forwarding criteria in the network. Mapping
of QCI values to DSCP is well understood and GSMA has defined standard
means of mapping between these scalars .
Primarily LTE offers two types of bearer: Guaranteed Bit rate bearer for
real time communication, e.g., Voice calls etc. and Non-Guaranteed bit
rate bearer, e.g., best effort traffic for web access etc. Packets
mapped to the same EPS bearer receive the same bearer level packet
forwarding treatment.The Web Real-Time communication (WebRTC) framework provides the
protocol building blocks to support direct, interactive, real-time
communication using audio, video, collaboration, games, etc., between
two peers' web-browsers. WebRTC application would use Interactive
Connectivity Establishment (ICE) protocol
for gathering candidates, prioritizing them, choosing default ones,
exchanging them with the remote party, pairing them and ordering them
into check lists. Once all of the above have been completed then the
participating ICE agents can begin a phase of connectivity checks and
eventually select the pair of candidates that will be used for real-time
communication.The following subsections describe different aspects of QoS for
WebRTC application in a 3GPP Network. in section 4.4
suggests to put interactive audio and interactive video over the same
5-tuple {dest addr, source addr, protocol, dest port, source port}.
Without special handling, the same QoS treatment will be applied to
the audio and video flows in the 5-tuple. One potential solution is
for the endpoints to set DSCP appropriately on a per-packet basis, as
proposed in . The packet
filters in UL TFT and DL TFT created for each media stream must also
include DSCP values, so that multiplexed upstream and downstream
traffic is separated and mapped to the correct bearer. That means that
there could be multiple bearers (packet filters) with the same 5-tuple
but with different DSCP values.In all cases where the browser multiplexes different priority media
flows on a single 5-tuple and RTP session, the browser should ensure
that the RTCP packets for each media flow receive the same or better
priority as the corresponding RTP packets. One way for the browser to
ensure this is to construct compound RTCP packets only with RTCP
packets of the same priority.RTCWEB is designed for Internet calling, so there are occasions
where the remote peer will not be on the same carrier's network, might
be on DSL, might be on a DOCSIS network. In that situation, the DSCP
bits might not be preserved across the remote peer's network, over the
Internet, and into the local carrier's network. When this occurs DSCP
markings set by the remote peer are likely to be modified by the
intervening network(s). Also, DSCP markings requested by a JS
application for upstream traffic may not survive on the path to the
radio modem because of OS limitations. Given the limitations described
above the following techniques can be used to solve the problem : in section 4.4
also proposes not to multiplex interactive audio and interactive
video over the same 5-tuple for compatibility with legacy systems.
If DSCP cannot be propagated to differentiate the required QoS
treatment of the traffic within the same 5-tuple then RTP
Multiplexing can be disabled by either the JS application or a
middle box involved in signaling. Even though DSCP markings are
frequently modified by intervening equipment, it is preferable to
set them appropriately at each browser to reduce the chance of
blocking of higher priority packets in common queues on the path
between each browser and their corresponding radio bearer(s).RTCWEB is pursuing a option where SSRC(s) for each flow be
explicitly signaled in the SDP, thus providing an other way of
identifying each flow (i.e., filtering on five-tuple and SSRC
value). If 3GPP agrees to extend the packet filter definition to
include RTP SSRC, then the UL and DL TFTs would not include any DSCP
values and the system would be able to assign each flow to the
appropriate bearer/QoS based solely on the assigned SSRC values
within the 5-tuple. This requires modification to the 3GPP
specifications to extend the TFT filtering mechanism to also support
matching on RTP SSRC values. With this extension, multiplexed audio
and video media streams using known RTP SSRC values can to be mapped
to different EPS bearers, thus receiving different packet forwarding
treatment.The TFT with RTP SSRC might not be applicable in some non-WebRTC
use cases and in some WebRTC use cases involving interoperation with
legacy peers. In some applications the SSRC value associated with a
particular media flow might change over time due to resource
switching. In any case where the SSRC value might not be known for
every media flow, a signaling level entity can either disable
multiplexing, assign only media flows that are intended to receive
the same QoS treatment to any one 5-tuple, establish another means
of determining the SSRC value for each flow so that the TFT can be
updated, or introduce a media gateway to map the SSRC value in each
flow to an explicitly signaled value. Any solution other than the
first (disabling multiplexing) requires further study.BUNDLE requires that each payload type (PT) value used across
media lines in a bundled group is unique, thus providing a way of
identifying the media flow(s) associated with an m line (i.e.,
filtering on five-tuple and PT values). If 3GPP agrees to extend the
packet filter definition to include RTP PT, then the system would be
able to assign the flow(s) associated with each m line to the
appropriate bearer/QoS based solely on the PT values within the RTP
header of packets in the 5-tuple. With this extension, multiplexed
audio and video media streams using known RTP PT values can to be
mapped to different EPS bearers, thus receiving different packet
forwarding treatment.Although this approach is applicable to any use of BUNDLE, there
are some limitations. Since a typical m line negotiates the use of
multiple payload type values, the TFTs for a typical m line are
likely to be more complex then those based on SSRC. Note that if
multiple media flows are associated with a particular m line then
the network could not distinguish among them and they would of
necessity receive the same QoS treatment. Most significantly, since
the PT field has a different meaning in RTCP packets, the TFTs for
RTP packets do not apply to RTCP and there is no information in the
RTCP packets to identify the priority treatment they should receive,
even if only RTCP packets of the same intended priority are
assembled in any one compound RTCP packet (assuming the DSCP value
cannot be trusted). The only solution in this case is to create a
TFT for all RTCP packets to assign them the highest priority
assigned to any of the RTP packets in the bundle. It is not
desirable to use higher priority bandwidth for lower priority RTCP
packets but it is less desirable to assign high priority RTCP
packets to a lower priority bearer under congestion. SSRC and PT
seem viable solutions and and could be used in future based on
further study.RTCWEB is pursuing designs which multiplex non-RTP flows with RTP
flows in the same 5-tuple. In particular for WebRTC, data channel
will use SCTP/DTLS/UDP, potentially on the same 5-tuple as SRTP/UDP.
Further a data channel could have multiple streams with different
priorities multiplexed within it as explained in section 5.1 of
and a mechanism
is required to identify and provide differential QoS per stream. If
DSCP marking set by the remote peer is preserved and the OS allows
setting per-packet DSCP then data channel can be multiplexed with
the media streams. However when DSCP is changed along the path or OS
does not allow setting per-packet DSCP then the following aspects
needs to be considered :Since SCTP stream IDs are encrypted, there is no way to enhance
the TFT definition to enable differential QoS treatment among the
multiple streams multiplexed within an single SCTP association on a
5-tuple. It is possible to provide the same treatment to all streams
within the SCTP association, using one of the following two
approaches: 1) after assigning traffic that matches on specific RTP
SSRC values within a 5-tuple to certain EPC bearers, assign the
remaining traffic within the 5-tuple to another EPC bearer/QoS; or
2) create new TFT extensions to allow identification of other
protocols and protocol-specific flow IDs to the extent this
information is available in unencrypted fields. Approach 1) is
available if only the RTP SSRC or RTP payload extension is defined
for TFTs. Approach 2) is for further study. The browser can still
provide differential treatment to streams with different priority
within the available bandwidth provided for a data channel among
packets it transmits within a 5-tuple.If the UE is using a Public WebRTC server then the UE can request
bearer resource modification for an E-UTRAN as explained in section
5.4.5 of . The procedure allows the UE
to request modification of bearer resources (e.g., allocation or
release of resources) for one traffic flow aggregate with a specific
QoS demand. Alternatively, the procedure allows the UE to request
modification of the packet filters used for an active traffic flow
aggregate, without changing QoS. If accepted by the network, the
request invokes either the Dedicated Bearer Activation Procedure or
the Bearer Modification Procedure.Problems with this approach are:When the UE initiates a call using WebRTC application and
candidate pairs are successfully nominated for each media stream
then WebRTC application should signal the packet filter, QCI and
GBR values for each media stream that would initiate bearer
resource modification or dedicated bearer activation. The WebRTC
application needs to be aware of the QCI/GBR values required for
the media streams.The WebRTC application requires API's to indicate to the
OS/Modem the packet filters for each media stream to be added,
modified or deleted. The WebRTC application would need API's to
indicate to the OS/Modem the QCI and GBR values for each of the
packet filters. The OS/Modem would in turn signal this information
to the 3GPP network that would result in either bearer resource
modification or creation using the Dedicated Bearer Activation
Procedure.WebRTC application may also need API's to trigger the
modification of the GBR value for existing packet filters.This problem is not unique to WebRTC and is applicable to any
other application that wants to trigger bearer resource modification
from the MN.Currently 3GPP networks prioritize flows by examining the SDP in
SIP signaling. As WebRTC also uses SDP in signaling to establish a
PeerConnection, similar mechanisms can be used to deploy a WebRTC
server in a 3GPP network.3GPP network cannot prioritize all a=candidate lines described
in until WebRTC server receives an
indication of the active media path from the controlling ICE
agent. WebRTC server is aware of the active media path only after
the controlling ICE endpoint follows the procedures in Section
11.1 of , specifically to send
updated offer if the candidates in the m and c lines for the media
stream (called the DEFAULT CANDIDATES) do not match ICE's SELECTED
CANDIDATES (also see Appendix B.9 of ).WebRTC server deployed in 3GPP network would need "Rx”
interface to the Policy and Charging Rules Function (PCRF). The
PCRF is the policy server in the EPC. WebRTC server would act as
"Application Function". Dynamic PCC (policy and charging control)
rules are derived within the PCRF from information supplied by the
AF (such as requested bandwidth for the 5-tuple, Application
Identity etc). PCRF forwards PCC rules for the media stream to the
Policy Charging and Enforcement Function (PCEF) for packet
scheduling, data packet (Diffserv) marking, etc., to allow QOS to
be provisioned in the EPC. Bearers would have be modified/created
for media streams, assigned and installed on the UE.TODO3GPP has a current work item on "Service Awareness and Privacy
Policies" that is chartered to add DPI-related extensions to the PCC
architecture . The (optional) DPI
entity in the EPC is called "Traffic Detection Function" (TDF), and it
performs application detection and reporting of detected application
and its service data flow description to the Policy Control and
Charging Rules Function (PCRF) for performing functions such as
traffic blocking, redirection, policing for selected flows.If UE is using Public WebRTC server thenThe session signaling between the WebRTC application running in
the browser and the web server could be using TLS. Moreover WebRTC
does not enforce a particular session signaling protocol to be
used, so network gateways in 3GPP network would fail to inspect
the signalling to identify the 5-tuple used for media stream and
thus fail to prioritize media traffic. Hence derived service
identification would not
succeed.Network Gateways by inspecting L7 traffic can only identify RTP
but fail to distinguish between IPTV vs. Multimedia, Gaming vs.
Voice Chat, Gaming vs. Voice Chat #2 etc as explained in section
3.1 - 3.3 of .The following section lists the potential problems If UE uses Public
WebRTC server :Given the exponential growth in the mobile data traffic, Mobile
Operators are looking for ways to offload some of the IP traffic flows
at the nearest access edge that has an Internet peering point. This
approach results in efficient usage of the mobile packet core and
helps lower the transport cost. Since Release 10, 3GPP starts
supporting of Selected IP Traffic Offload (SIPTO) function defined in
,. The
SIPTO function allows an operator to offload certain types of traffic
at a network node close to the UE's point of attachment to the access
network. Limited Mobility support available with SIPTO is explained in
section 2.3.3 of .If SIPTO is carried out in a Traffic offload Function (TOF) entity
located at the interface of the Radio Access Network i.e. In the path
between the Radio stations and the Mobile Gateway (MGW). The TOF
decides which traffic to offload and enforces NAT for that traffic.
The deployment of a TOF is totally transparent for the UE and hence
does not know which traffic is subject to TOF (NATed at the TOF) and
which traffic is processed by the MGW.The problem with WebRTC application in such network is thatTOF is not aware of the 5-tuple that will used for media and
data channels. ICE agent would gather server reflexive candidates
using STUN and relayed candidates are obtained through TURN. If
STUN messages are offloaded at TOF then UE would learn the
External IP Address/Port provided by the NAT at TOF. Similarly ICE
connectivity checks could also be offloaded at TOF. If UE roams,
though host candidate addresses may not change but NAT will change
resulting in failure to reach the remote peer for the existing
media and data channels. If the media and data channels are
offloaded at the TOF then UE Mobility would result in disruption
of media and data channel traffic.If UE is using local relayed candidate to reach the remote peer
and roams out of the coverage of RNC/HNB GW then NAT between UE
and TURN server changes, so UE cannot use the previous TURN
allocations and fail to reach the remote peer using local relayed
candidate. can be used in
such scenarios to provide media session mobility.Proxy Mobile IPv6 (or PMIPv6, or PMIP) is a network-based mobility
management protocol specified in .
Network-based mobility management enables the same functionality as
Mobile IP, without any modifications to the host's Protocol stack.
With PMIP the host can change its point-of-attachment to the Internet
without changing its IP address. defines a way to
signal the Traffic Offload capability of a Mobile Access Gateway (MAG)
to the Local Mobility Anchor (LMA) in Proxy Mobile IP Networks. Mobile
access gateway has the ability to offload some of the IPv4 traffic
flows based on the traffic selectors it receives from the local
mobility anchor. Using IP Traffic Offload Selector option mobile access
gateway will negotiate IP Flows that can be offloaded to the local
access network.The problem with WebRTC application in such network is thatMAG and LMA are not aware of the 5-tuple that will used for
media and data channels. If STUN messages are offloaded at local
access network then UE would learn the External IP Address/Port
provided by the NAT at local access network. Similarly ICE
connectivity checks could also be offloaded at local access
network. If UE roams out of the coverage of Local Access Network
though host candidate addresses may not change but NAT will change
resulting in failure to reach the remote peer for the existing
media and data channels. If the media and data channels are
offloaded at the local access network then UE Mobility will result
in disruption of media and data channel traffic.If UE is using local relayed candidate to reach the remote peer
and roams out of the coverage of Local Access Network then NAT
between UE and TURN server changes, so UE cannot use the previous
TURN allocations and fail to reach the remote peer using local
relayed candidate. can be used in
such scenarios to provide media session mobility.The
proposes extensions to Prefix Information Option with a mobility flag bit. This would allow
for network based mobility solutions, such as Proxy Mobile IPv6 or GTP to
explicitly indicate that their prefixes have mobility and therefore,
the UE IP stack can make an educated selection between prefixes that
have mobility and those that do not. If WebRTC application requires
mobility for media streams then it can pick source addresses generated
from prefixes with 'M' Flag set to 1 in Prefix Information Option or
pick IPv6 prefixes without 'M' flag set to 1 if both the endpoints
support . If WebRTC
application requires mobility for media streams and the remote peer
does not support
then it can still pick prefixes without 'M' flag set to 1 when it uses
relayed local candidates for the media streams and TURN supports
Mobility as explained in section 5 of .TODO: This section needs more details to be added for WebRTC
application to pick suitable prefix.This document does not define an architecture nor a protocol; as such
it does not raise any security concern.This document does not require any action from IANA.Authors would like to thank Dan Wing, Basavraj Patil, Magnus
Westerlund, Markus Isomaki, Cullen Jennings, Harold Lassers, Thomas
Anderson, Charles Eckel, Subha Dhesikan , Parthasarathi R, Reinaldo
Penno, Prashanth Patil for their comments and review.End-to-End Quality of Service (QoS) Concept and Architecture,
Release 10, 3GPP TS 23.207, V10.0.0 (2011- 03)"General Packet Radio Service (GPRS); Service description;
Stage 2", June 2012.General Packet Radio Service (GPRS) enhancements for Evolved
Universal Terrestrial Radio Access Network (E- UTRAN) access
(Release 11), 3GPP TS 23.401, V11.2.0 (2012- 06).3GPP, "3GPP Evolved Packet System (EPS); Evolved General
Packet Radio Service (GPRS) Tunnelling Protocol for Control plane
(GTPv2-C)", 3GPP TS 29.060 8.11.0, December 2010.3GPP, "Policy and charging control architecture", 3GPP TS
23.203 10.5.0, December 2011.Inter-Service Provider Backbone Guidelines 5.0, 22 December
2010TODO : In Windows setsockopt is completely disabled. See
Knowledge Base Article http://support.microsoft.com/kb/248611.DSCP is supported and user settable on Symbian S60, Linux, MacOS
X.