Download Out-of-band Signalling Channel for Efficient Multicast Service Delivery in

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Wireless security wikipedia , lookup

IEEE 802.1aq wikipedia , lookup

Distributed firewall wikipedia , lookup

Computer network wikipedia , lookup

Network tap wikipedia , lookup

Cracking of wireless networks wikipedia , lookup

Cellular network wikipedia , lookup

Airborne Networking wikipedia , lookup

Piggybacking (Internet access) wikipedia , lookup

Zero-configuration networking wikipedia , lookup

Transcript
IWS 2005 / WPMC’05 - Aalborg, Denmark
Out-of-band Signalling Channel for Efficient Multicast Service Delivery in
Heterogeneous Wireless Networks
Alexander Gluhak, Klaus Moessner, Rahim Tafazolli
Centre for Communication Systems Research
Mobile Communications Research Group
The University of Surrey
Guildford, GU2 7XH, UK
Email: {a.gluhak,k.moessner, r.tafazolli}@surrey.ac.uk
The reminder of the paper is structured as follows. Section 2
outlines the application scenario, motivating the employment of
the proposed MSC. In section 3 the MSC is presented, providing
an analysis of the employment gains of the MSC. Furthermore the
proposed scheme for receiver subset addressing on the MSC is
described in more detail. Then guidelines for signalling network
selection are provided in section 4. Concluding remarks are given
in section 5 outlining future research directions.
Abstract— Multicast delivery in heterogeneous wireless networks requires careful coordination, in order to take full
advantage of the resources such an interworking network
environment can offer. Such coordination may demand interworking signalling from coordinating network entities to
receivers of a multicast service. Scalable delivery of interworking signalling to receivers of multicast user services is
of great importance, since the numbers of receivers may be
very large. This paper therefore investigates the use of a
multicast signalling channel (MSC) to carry such interworking
signalling in a scalable manner. Analytical evaluations are
performed, comparing required signalling load on the MSC to
unicast signalling. In order to further increase the efficiency
of the MSC a novel approach is proposed that allows efficient
addressing of a subset of receivers within a multicast group.
Furthermore guidelines for the selection of a suitable signalling
network carrying the MSC are provided.
2. APPLICATIONS FOR MULTICAST CONTROL
SIGNALLING
Previous work within the MIRAI [2] project, has proposed an
out-of-band signalling mechanisms and network entities to facilitate seamless network interworking [3]. The control signalling, also
called basic access signalling (BAS) is used for user registration,
radio access network discovery, location update, session setup, and
media handover. Such signalling can be provided over a dedicated
basic access network (BAN), used only for the signalling purposes,
or over one of the available radio access networks. The proposed
mechanisms, however, mainly targeted unicast communication.
Our work draws on the concepts and experiences of MIRAI
and aims to provide an efficient signalling solution for coordinated
multicast service delivery in HWN. Previous work in the Core 3
research program of Mobile VCE [5] has therefore proposed and
investigated a more network-centric approach [4], which allows the
coordination of multicast user services by (possibly distributed)
network management entities. Using scenario based analysis the
required signalling for advanced interworking functionality such
as dynamic network selection or load balancing have been studied.
It has been found that in the case of multicast service delivery in
HWN, the same interworking signalling information may have to
be delivered to multiple receivers. The following signalling applications involving groups of receivers have thus been identified:
Key words: Multicast Service Delivery, Control Signalling,
Heterogeneous Wireless Networks
1. INTRODUCTION
While looking for different ways to improve the efficiency of service delivery, mobile network operators have recently discovered
multicast delivery as promising solution in mobile networks and
are currently standardising Multimedia Broadcast and Multicast
Services [1] in UMTS Release 6. Furthermore in order to offer
superior services to their mobile customers, operators with different
access network technologies, e.g. DVB and WLAN, are likely
to cooperate in a future network scenario. Such a heterogeneous
wireless network (HWN) environment, however, requires new
mechanisms to facilitate efficient network interworking for multicast service delivery.
This paper investigates the use of a multicast signalling (MSC)
channel as a scalable approach for delivering downlink signalling
for network interworking support to groups of heterogeneous
receivers. Our analysis reveals two main issues that are crucial for
the design of such an MSC, namely the selection of an appropriate
signalling network and effective addressing of subset of receivers
subscribed to the signalling channel. We give guidelines for the
selection of an appropriate signalling network and propose a novel
efficient mechanism for addressing a subgroup of receivers within a
multicast group. The proposed mechanism minimises the required
signalling load by allowing an aggregation of receivers, which is
based on context information receivers may have in common.
Copyright  2005 WPMC
Masugi Inoue
National Institute of Information
and Communications Technology
3-4 Hikarino-oka, Yokosuka
Kanagawa, 239-0847 Japan
Email: [email protected]
•
•
1444
Network-initiated establishment of multicast bearers.
In order to allow advanced interworking functionality such
as dynamic access network and bearer selection, network
operators need to initiate the establishment of suitable multicast bearers. Intelligent algorithms [6] could select suitable
bearer paths, e.g. according to availability of current network
resources or terminal capabilities and then trigger the establishment of those at interested receiver groups.
Vertical network handoff for groups of receivers.
During a multicast session conditions may arise where a
18-22 September 2005
IWS 2005 / WPMC’05 - Aalborg, Denmark
•
•
change of current access network may be required. An example of a network initiated vertical handoff would be a load
balancing situation: In order to free up their own network
resources, an operator may decide to handover a group of
receivers in several cells of an access network to an alternative
access network of an interworking operator. Thus signalling
has to be delivered to the respective receivers in order to
initiate a change of the delivery network.
Flow handoff for groups of receivers.
Often a change of access network may lead to a change
of the delivery characteristics e.g. bandwidth availability. In
such cases the flow within a multicast session may change,
e.g., different QoS of a media flow or addition/removal of
media flows. Furthermore to increase the delivery flexibility,
e.g. in emergency situations, receivers may be temporarily
downgraded to a lower rate flow in order to free up network
resources or may be upgraded if sufficient network resources
are available.
Radio access network discovery: Network assisted radio access network discovery allows terminals to have knowledge of
available access networks, without the need for frequent scans
on all network interfaces. Thus significant power saving can
be achieved [7]. Groups of receivers at a similar location can
be provided with the same network discovery information.
IP Header (Unicast)
(a)
IP Header (Multicast)
Addressing Expr.
Signalling Payload
(b)
Figure 1.
message.
Format of a unicast(a) and a multicast(b) signalling
as addition to the signalling message. The addressing expression
allows to identify the subset of receivers, for which the signalling
message is actually intended. Receivers evaluate the addressing
expression above the network layer, and based on the outcome of
the assessment, accept or discard the signalling message. Figure 1
shows the format of a control message sent via unicast (a) and a
multicast message (b) on the MSC.
3.2. Analytical Evaluation
In order to employ the proposed MSC successfully, its advantages and disadvantages need to be fully understood. This section
therefore presents an analytical evaluation of the signalling load of
the MSC approach, by comparing the signalling load requirements
on the MSC to unicast delivery. The analysis considers only the
bandwidth requirements on the last hop wireless link and neglects
the core network resources. This assumption can be justified, since
the wireless resources usually represent the bottle neck within a
wireless network.
Parameters for the analytical comparison of required signalling
load for both cases are described in the following: let Cuni be the
signalling cost for message delivery in unicast case and Cmulti
be the respective signalling cost for delivery over the multicast
signalling channel. Furthermore let Nuni denote the number of
bits for an IP header in the unicast case, Nmulti the number of
bits for an IP header in the multicast case, Nexp the size of the
addressing expression required for receiver subset addressing on
the MSC and Nmsg the number of bits of the actual higher layer
message payload. Sending a message to r receivers, the following
signalling costs can be identified for the unicast delivery case:
In order to provide the required control signalling in HWN to a
potentially large group of receivers, a scalable way of delivery is
essential. In the next section we explore the deployment of a MSC
to carry control signalling from coordinating network entities to
groups of receivers.
3. MULTICAST SIGNALLING CHANNEL
3.1. Principles
The key idea for the deployment of a MSC is to reduce the
number of signalling messages and hence the signalling load in the
network, when the same signalling message needs to be delivered
to a group of receivers. Instead of delivering a signalling message
via unicast individually to all affected receivers, a single message
would be send to a specific multicast group address, identifying
the MSC. Receivers, which need to receive the signalling message
would subscribe to the multicast group address of the MSC over
which the message is sent.
Ideally only receivers, for which a control signalling message
is intended, should be subscribed to the MSC at the time the
signalling message is sent. While this concept may work perfect
in theory, it exhibits some practical problems. In most cases a
control message will not be intended for all receivers of a multicast
user service. Rather a signalling message, such as a request for
vertical handoff would target only a subset of receivers, e.g. only
receivers which are capable of establishing a multicast bearer in
the new access network. This would require a way to inform the
receivers to subscribe or unsubscribe from the MSC, whenever a
new message needs to be sent. While such notifications will add
to the required signalling load, frequent ’joins’ and ’leaves’ of
receivers to multicast groups will further increase the signalling
load, thus reducing the foreseen benefits of the MSC.
It is therefore more realistic to assume that all receivers of a
particular multicast service are subscribed to the same MSC for
the lifetime of the session. In case of access network discovery,
receivers at certain geographic areas receive their information from
the same MSC. Within the MSC an addressing expression is used
Copyright  2005 WPMC
Signalling Payload
Cuni = r ∗ (Nuni + Nmsg )
(1)
While the cost for delivering a control signalling message via
unicast is directly proportional to the number of receivers, the
signalling cost for a multicast message is independent of the
number of receivers as seen in equation 2. In contrast the signalling
cost for delivering a message via multicast is proportional to the
number of cells m, in which wireless access network resources are
utilised.
Cmulti = m ∗ (Nmulti + Nexp + Nmsg )
(2)
Assuming that the signalling network used for the MSC consists
of multiple cells, a multicast bearer needs to be established in
all cells, which host receivers that are subscribed to the MSC. A
signalling message is sent in each of the cells, regardless of the
presence of receivers from the subset, for which the messages was
actually intended.
A main difficulty is thus to find the break-even point, which
justifies the utilisation of a MSC. Comparing the signalling load
of a single message delivery, the use of an MSC can be justified
if the following applies:
1445
18-22 September 2005
IWS 2005 / WPMC’05 - Aalborg, Denmark
Control Signalling Load − Payload: 1000 bits
Control Signalling Load − Payload : 1000 bits − Addressing Expression: 1000 bits
8
30
Unicast
Unicast
25
6
Signalling Load (kBytes)
Signalling Load (kBytes)
7
5
4
3
MSC
2
20
15
10
MSC
5
1
0
10000
0
50
8000
40
50
6000
40
200
30
30
4000
20
2000
Addressing Expression (Bytes)
0
100
10
10
0
150
20
Number of Cells
Number of Receivers
Figure 2. Signalling load for unicast and multicast delivery for
different addressing expression sizes in a single cell.
50
0
0
Number of Receivers
Figure 3. Unicast vs multicast delivery of a control signalling
message if receivers are distributed in multiple cells.
3.3. Context Information Based Receiver Subset Addressing
Cmulti ≤ Cuni
The signalling expression on the MSC is used to identify a
receiver subset, in case the control message does not apply to all
subscribed receivers. In order to minimise the signalling load on
the MSC, the required addressing expression should be kept as
small as possible.
The simplest way to identify a receiver subset is to explicitly
encode unique identifiers of each receiver in the addressing expression. Examples of unique identifiers are unicast IP address, as
used in the XCast approach [8], or host identity [9]. The approach,
however, does not scale well, if the signalling message needs to
address a large number of receivers. Assuming 10.000 receivers
and the use of a IPv4 unicast address the explicit addressing
expression would be as large as 320kBits! Therefore more powerful
expressions to aggregate addressing for large subset of receivers
are necessary.
Analysing the signalling application scenarios, following important observation has been made: In many cases the control
signalling targets a group of users with common context e.g.
all receivers in a certain area/cells of a network or all receivers
currently receiving a flow with the same quality of service. This
has motivated to propose a receiver aggregation mechanism for
the MSC that is based on receiver context information. A subset of
receivers on the MSC is described by context information that those
receivers have in common. The requirement for such a mechanism
to work is that the network entity, sending the control messages via
the MSC has access to the context information of these receivers.
Moreover receivers need to evaluate the specification of the context
information provided in the addressing expression and be able to
infer, wether or not the specification applies to them. Useful context
information, which can be used for such address aggregation has
been identified and is described in the following:
• Receiver location:
Often control signalling will target a set of receivers at a
geographic location. Geographic locations can be described
logically by specifying network and cells or by GPS information. For example load balancing will try to free up
resources in a cell or cell cluster of a network by switching
the multicast bearer to another access network with more
available resources. Users in a cell cluster could be easily
(3)
Substituting (3) by (2) and (1), and with Nmulti = Nuni = Nip
the threshold can be rewritten showing the multicast dependent
parameters on the left and the unicast dependent on the right hand
side:
m∗(
Nexp
+ 1) ≤ r
Nip + Nmsg
(4)
Figure 2 shows the signalling load required for delivering a
single control message in a single cell as a function of the number
of targeted receivers and the size of required addressing expression.
A signalling payload of 1000 bits as well as the use of IPv4 headers
has been assumed. The figure depicts the signalling load required
for the unicast case as well as for the multicast delivery using the
MSC. When an addressing expression of 1000 bit size is used to
identify a receiver subset, the use of an MSC becomes already
efficient, if two or more receivers are targeted by a signalling
message. Even for large addressing expression sizes such as 10000
bits (ten times the size of the signalling payload), the use of an
MSC can be justified if 10 or more receivers are targeted by a
control message. In order to allow efficient delivery also to a small
number of targeted receivers, the addressing expression should be
kept as small as possible. Therefore a novel technique for efficient
receiver subset addressing on the MSC is presented in section 3-3.
Most likely not all of the receivers, which are subscribed to
the MSC will be located in a single cell of the utilised signalling
network. Receivers will be scattered in several cells, for which
network resources need to be utilised. Figure 3 shows the overall
signalling load for a control message as a function of the number of
targeted receivers and the number of cells in which this receivers
are distributed. A payload size of 1,000 bits and an addressing
expression size of 1,000 bits have been used in the example. At
least 94 receivers need to be targeted in the case of 50 cells, in
order to gain an advantage by employing a MSC. Therefore the
use of an access network with large cell coverage is advantageous
if receivers are geographically wide distributed. Section 4 provides
a more detailed discussion on signalling network selection.
Copyright  2005 WPMC
1446
18-22 September 2005
IWS 2005 / WPMC’05 - Aalborg, Denmark
•
•
•
Table 1. Context information types, associated attribute types and
example values.
aggregated by an expression such as ’all receivers in cells
4,5,6,7 of a UMTS network’.
Terminal capabilities:
Terminal capabilities are attributes that can be useful to
express commonness among users. Terminal capabilities include available network access interfaces (NAI), maximum
supported QoS for a connection on a NAI, available memory,
software and codecs etc.. A control message initiating a
vertical handoff should move all receivers with an appropriate
NAI to a different access network. An expression such as
’all receivers with NAI type UMTS’ represents a short but
powerful aggregation for such a use case.
Receiver preference:
As with terminal capabilities, receiver preferences are attributes, which can be used to identify a subset of receivers.
Preferences can be expressed in ’delivery network’ or the
’QoS’ for the service flows. A control message using preferred
access network as a receiver aggregation could initiate the
establishment of a multicast bearer service on the respective
access network for all receivers, which have their preferences
in common.
Communication context:
Communication context describes the multicast bearers and
flows of a receiver that are currently associated with the
reception of a multicast service. Communication context can
be described by delivery network, multicast group address,
source address, QoS of received flows and source and destination ports. Receivers are aware of their communication context
and the information can be used to efficiently aggregate
receivers, e.g. receivers subscribed to a common multicast
bearer or receiving the same service flow. Note that unlike
the previously mentioned context information, communication
context does not exist before a communication is established.
Therefore communication context can be only used for addressing expressions of control messages, which are sent
during a session.
Attribute Type
Example Value
Location
Network
Cell
GPS
Terminal
Capabilities
Network Interface
Supported QoS
UMTS,DVB
256k
Service
Preferences
Delivery Network
Desired QoS
DVB
384k
Communication
Context
Delivery Network
Current QoS
Multicast Addr
Source Addr
Port
WLAN
384k
239.27.0.8
123.44.32.2
1500
DVB
1,2,5
◦
N49 00’ 32´’. . .
to maximise the flexibility in defining a receiver subset with an
addressing expression, a combined addressing expression using
explicit addressing and context information based aggregation can
be used.
We have implemented a prototype of the MSC utilising the
proposed addressing scheme for receiver subset addressing and
performed successful tests in a network environment. In the initial
version, XML has been used to encode addressing expressions.
XML has been chosen due to its good human readability and
extensibility by new context information types. Figure 5 shows
an example of an XML encoded addressing expression.
During the experiments with the prototype it was found that the
use of XML adds considerable overhead to the context information
based addressing expression. The overhead comes mainly from
the tags, which defined the structure and nature of the addressing
information. Since the addressing expression size is critical for the
performance of the MSC, improvements in the representation of
contextual information will have to be made, before deployment
in a real world system. One way would be the replacement of
the currently used tags by shorter ones at costs of the readability,
which would not matter once the experimental phase is concluded.
A more effective alternative would be the definition of proprietary
binary encoded information elements.
The above list of context information can be easily extended as
new useful context information for aggregation becomes available.
Within the addressing expression, context information is expressed
as a key-value pair, with the type of context being the key, and
a context attribute representing the value. Context attributes can
be further composed of attribute-types and associated attributevalues. For example the context type ’location’ can be expressed
by attribute types ’network’ and ’cell’, with ’UMTS’ and ’7,8,9’
being examples of the respective attribute-values. Two expressions
of context information can be combined by logical operators
to characterise a receiver subset more specifically. Furthermore
the notion of a compound attribute is introduced, which allows
arbitrary combination of key-value pairs with logical operators.
Table 1 shows an overview of commonly used context types,
associated attribute-types and typical values.
4. SIGNALLING NETWORK SELECTION
The selection of an appropriate signalling network is an important means to reduce the overall signalling load required for
the delivery of a message on the MSC. A signalling message,
which is sent on the MSC is transmitted in every cell in which
the MSC is established. Thus available bandwidth and hence radio
resources are utilised in each of those cells. Although all receivers
of a multicast user service would subscribe to a respective MSC,
Figure 4 shows an example of a context information based
addressing expression. The expression describes all receivers with
a DVB network interface present in cells 6,7 and 8 of a UMTS
network. In a load balancing scenario a control message with
this addressing expression could thus initiate a vertical handoff
to a DVB network to free up resources in cells 6,7,8 of the
UMTS network. As demonstrated in the example, a receiver subset
can be accurately identified with a relatively short expression,
instead of using hundreds or more explicit addresses. In order
Copyright  2005 WPMC
Context
Information Type
{ [’Location’:([’Network’:’UMTS’] & [’Cells’:’6,7,8’])] &
[’Interface’:’DVB’] }
Figure 4. Example of a context information based addressing
expression.
1447
18-22 September 2005
IWS 2005 / WPMC’05 - Aalborg, Denmark
5. CONCLUSIONS
<address>
<compound>
<location>
<network>UMTS</network>
<cells>1,2,4,5</cells>
</location>
<operator>AND</operator>
<capabilities>
<network>DVB</network>
</capabilities>
</compound>
<operator>AND</operator>
<preference>
<network>DVB</network>
</preference>
</address>
Figure 5.
In this paper a multicast signalling channel has been proposed to
efficiently carry out-of-band interworking signalling to groups of
receivers for coordinated multicast service provisioning in HWN.
The selection of an appropriate signalling network as well as
efficient mechanisms for addressing a subset of receivers have
been identified as crucial design issues for such a signalling
channel. While giving guidelines for the selection of a suitable
signalling network, we also propose a novel efficient mechanism
for addressing a subset of receivers within a multicast group,
by aggregation of receivers based on context information they
have in common. Despite the potential signalling performance
gains of the MSC, situation may occur, where unicast signalling
would be preferable, e.g. very small subset of receivers to be
targeted by a control message. In such cases, a hybrid solution
may be employed, which could switch between multicast and
unicast signalling delivery based on a threshold. Such mechanisms
however require further investigation.
Example of an XML encoded addressing expression.
ACKNOWLEDGMENT
The work reported in this paper has been part of a collaboration
of the Virtual Centre of Excellence in Mobile & Personal Communications, Mobile VCE, www.mobilevce.co.uk and National
Institute for Information and Communication Technolgy (NICT).
This collaboration has been made possible by Japanese Society for
Promotion of Science (JSPS), whose funding support is gratefully
acknowledged.
a signalling message would often target only a subset of those
receivers. In the worst case this subset of receivers is located in
a single cell, thus the message is unnecessarily transmitted in the
remaining other cells. In contrast no resources would be wasted
if receivers of the destined subset are present in each of the cells,
where the MSC is established.
REFERENCES
In order to utilise radio resource efficiently, the area covered
by the cells of the signalling network carrying the MSC should
approximately match the area occupied by the receivers subscribed
to the MSC. Ideally a signalling network providing the MSC would
cover all receivers subscribed to that channel by a single cell. That
way the receivers could be reached by a single transmission, utilising network resources only in a single cell. Therefore networks
employing larger cell structures such as broadcasting networks, e.g.
DVB, would be suitable candidates for carrying an MSC. The fewer
cells are needed to provide the MSC to the receiver group, the less
overall network resources are claimed for the transmission of a
message. However networks with a large cell sizes do not provide a
good granularity. In some cases networks with hierarchical network
structure, e.g. employing both micro and macro cell overlays would
provide the highest flexibility. Thus micro and macro cells could
be combined for providing the MSC in the required area, without
unnecessarily trading in radio resource and coverage efficiency.
[1] T. S. G. Services and S. Aspects, 3GPP TS 22.246 V6.2.0:
Multimedia Broadcast/Multicast Services (MBMS) user services; Stage 1 (Release 6). 3rd Generation Partnership Project
3GPP, Sept. 2004.
[2] G. Wu, M. Mizuno, and P. Havinga, “Mirai architecture for
heterogenoeus networks.” IEEE Communications Magazine,
pp. 126–134, Feb. 2002.
[3] M. Inoue, K. Mahmud, H. Murakami, M. Hasegawa, and
H. Morikawa, “Novel out-of-band signalling for seamless
interworking between heterogeneous networks.” IEEE Wireless
Communication Magazine, pp. 56–63, Apr. 2004.
[4] A.Gluhak, K.Moessner, and R.Tafazolli, “Controlled multicast
service delivery in heterogeneous wireless networks.” in Proceedings of the IEICE General Conference, Osaka, Japan.
IEICE, Mar. 21–24 2005.
[5] “Mobile Virtual Centre of Excellence, www.mobilevce.com.”
[6] A.Gluhak, K.Chew, K.Moessner, and R.Tafazolli, “Multicast
bearer selection in hetergenoues wireless networks.” in Accepted for publication by the International Conference on
Communication, ICC 2005 , Seoul, Korea. IEEE, May 16–20
2005.
[7] K. Mahmud, M. Inoue, H. Murakami, M. Hasegawa, and
H. Morikawa, “Energy consumption meassurement of wireless
interfaces in multi-service user terminals in heterogeneous
wireless networks.” IEICE Transactions on Communications,
pp. 1097–1109, Mar. 2005.
[8] R. Bovie and et al., draft-ooms-basic-spec-00.txt: Explicit
Multicast (XCast) Basic Specification. Internet Engineering
Task Force IETF, Dec. 2000.
[9] R.Moskowitz, P. Nikander, P.Jokela, and T. Hendersson, draftmoskowitz-hip-08.txt: Host Identity Protocol(HIP). Internet
Engineering Task Force IETF, Oct. 2003.
Finally, problems selecting a suitable signalling network may
arise from the heterogeneity of receivers. The above discussion
was based on the assumption that the MSC would be provided
to all receivers via a a common signalling network. This however
may not be feasible in a HWN environment, since receivers may
not have the same network interfaces. In some cases there may
not be a common access network on which all receivers could
be reached. Also a common access network may not always be
the best choice to carry a MSC. In such cases a combination of
two or more access networks may be required to carry the MSC.
Furthermore in some cases it may be feasible to provide a hybrid
signalling solution consisting of multicast and unicast delivery e.g.
by serving the majority of receivers via an MSC on a common
signalling network and providing the message to the remaining
receivers via unicast.
Copyright  2005 WPMC
1448
18-22 September 2005