Celtic project cp5-013 march ml architecture for multiplay services

Дата канвертавання24.04.2016
Памер385.85 Kb.
  1   2   3   4   5   6   7   8   9   ...   16

Celtic project CP5-013

MARCH - ML architecture for multiplay services

Deliverable D1.3

Final report

Issue date: 07.09.2011

Version: 1.0

Security group: Public

Authors: Terje Tjelta (ed), Telenor
Name, partner (one person or more per partner)

Executive summary

Multilink requirements

Multilink operation must be an enhancement to deployed networks and made available where favourable, based on cost versus benefit considerations. Thus, services and applications designed for multilink operation must be designed to be accessible in either mode; the multilink mode providing better functionality and service quality at an acceptable cost.

Networks considered are mobile (e.g., 3G, 4G, WiMAX), fixed (e.g., Wi-Fi, WiMAX), and broadcast (e.g., DVB). The characteristics and capabilities of each network must be understood in order to define protocols and procedures for efficiently deliver over these links. The traffic is IP based causing special implications for unidirectional broadcast networks.

Timing relations are critical where multiple links are used to increase the service bandwidth. If multiple components of a compound service are delivered using separate links timing may be more relaxed. It follows that intermediate storage will applied to compensate for differences in transmission path delays and processing time and latency.

Receiver terminals must be suitably equipped for multilink operations. As a rule some negotiation between terminal, network control and content/service provider will be necessary to ascertain the feasibility of a multilink service session. This will determine what links that may be used to reach the terminal, the bandwidth requirements, and how the service/application may be adapted to the requested multilink session. It is important that such negotiation shall not seriously add to the access time of the wanted service. To the largest possible extent the user shall be unaware whether multilink is in operation or not, conditioned by a general user acknowledge that the extra cost incurred from multilink usage is accepted.

The mobility of multilink connected terminals requires that measures are taken to hand traffic to a different link. If a replacement link is not available one option is to modify the service (if appropriate) to suit the new network conditions.

The control regime requires that interfaces be defined for all relevant network components such that control and monitoring data may be interchanged and made available to the party responsible for overall multilink administration. Network operators must accept openness to their network interfaces such that multilink operation control can be delegated to a common agreed actor. Likewise, service and application providers must be aware of network multilink options and limitations. It is important to use standardised techniques and solutions to ensure future compatibility and extendibility.

Multilink functionality must to the largest possible extent be enabled through software upgrades of existing functional units to avoid large infrastructure investments.

Multilink scenarios

The aim was to identify scenarios beneficial to all stakeholders such as service and network providers, and end users. Service scenarios and network scenarios are both addressed as two different scopes or characteristics of multilink solutions. MARCH looked at a scenario scope five years ahead. Both fixed and wireless telecommunication networks and broadcast networks are addressed. The multilink capabilities can be exploited differently depending on the type of services to consider, and to whom and in which area they are provided.

Several advantages are common for many multilink scenarios. Aggregating the capacity of many links and networks obviously leads to a higher available overall bit rate. Improved service quality can be obtained through either increasing the useful bit rate, or by using the extra capacity for better error protection. Balancing network load by transferring load from one link to another when required enables more stable network operation.

Multilink scenarios may open business opportunities for both new and existing actors. Also actors not taking an active role in this market may experience increased revenues due to increased demand for content, and increased traffic volume in existing networks.

The research work on scenarios also revealed a need to identify roles and actors in a more defined manner, and also to introduce new roles, such as mediators, to enable and sell multilink services.

The scenarios used for economic evaluation have been selected from the set of scenarios presented in Deliverable D2.2. The qualitative analysis has taken a set of architectural elements from Deliverable D4.1. The different business models are motivated and driven by different factors like new actors in the market, network bandwidth aggregation, applications or type of customer. Several actors may be involved in the value chain of a complete multilink service, and that there is a clear potential for positive business. However, there are also many risks in the business potential for certain actors.

The quantitative analyses provide results for four scenarios:

  • Broadband services combining mobile 3G and Wi-Fi

  • Broadband services combining satellite DVB-S (DVB-RCS) and WiMAX

  • Security fleet management deploying all available networks

  • Campus 3D deploying all available networks

Part of the first scenario is presented in some detail below, while the others are simply summarised. For details see Deliverable D3.3. The results provides an early approach to the costs for these scenarios, considering that market assumptions are uncertain guesses and that the technological market tendencies are unclear for the nearest coming years.

All four scenarios are shown to be positive after two or three years. The potential may be significant in economic terms, and may offer interesting business if the rest of the conditions are right. However, care should be shown as there are considerable uncertainties with this type of analysis.

Broadband market, consumer spending, and forecast models

MARCH first worked out forecasts for subscription penetration and then traffic forecasts per subscription in the busy hour was estimated. These forecasts together enable total traffic forecasts during the busy hour which is the key information for planning and dimensioning of the networks.

Short-term and long-term traffic forecasts are crucial for network planning, dimensioning and for long-term strategies. However, there are definite uncertainties in the traffic forecasts which in turn may create traffic bottlenecks and congestions in the networks.

In addition there are already traffic bottlenecks in the broadband mobile network. Especially limited sector capacity at the base stations and limited backhaul capacity creates traffic problems in the busy hour. So far the problems are partly solved by introduction of traffic volume cap of 5 -10 GB per month for a subscription.

The long-term traffic forecasts both for the fixed broadband network and for the mobile broadband network in the busy hour show an explosive growth. For instance the mobile broadband traffic in 2015 is forecasted to increase 36 times the 2008 traffic level and 16 times the 2009 traffic level. In addition the fixed broadband traffic is estimated to be 19 times larger than the mobile broadband in 2009, while the proportion is reduced to 6 times larger in 2015.

However, it is important to underline that there are significant uncertainties in these forecasts. The continuous evolution of mobile broadband technologies may elevate the traffic forecasts. See Deliverable D3.4 for the full report on market development.

MARCH has examined revenue streams and pricing models for content such as video, to justify if there will be willingness to pay for digital services, Deliverable D3.5. The analysis was performed on four levels:

  • Discussed the market for Internet access and content services as a two-sided market

  • Identified potential digital revenue streams between end-user and content provider

  • Documented household spending in general, and on content and telecommunications

  • Developed forecasting models for spending and possible future substitution

For a two-sided market, i.e., end user and content provider, it is difficult to identify criteria for the right balance between them, but one general rule has emerged. It is the side that “cares” more about the other side that should pay more, all else equal. In addition end-user’s demand for content variety in time and scope imply that content providers pay the platform. For a shift in market balance in the two-sided market it is expected that the content providers “care” more about reaching the customers than vice versa, and the end-users will be more interested in varied offerings over time and scope. The current vast volumes of information and actors on the Internet are indications of customer preferences for variance. Today – opposed to the early days of Internet – providers of content, business in general and public are critically dependent on online presence and reaching their customers. Hence content providers might by now be more willing to pay the Internet access provider for access to end-users and related services, and in parallel establish direct customer relationships to the end-user.

Based on the assumption of a future substitution between traditional content expenditure and fixed/mobile broadband spending (including content and services) MARCH suggests a set of forecasting models:

  • Fixed broadband penetration forecasting models and forecasts

  • Mobile broadband penetration forecasting models and forecasts

  • Forecasts/forecasting methods for expenditure categories

  • Forecasting models for part (fraction) of the expenditure substituted by telecommunications

The forecasts for the first three listed models are quantified. The forecasting model for the fraction of the expenditure substituted by telecommunications is only given as an example. It is necessary to quantify the possible fraction for substitution in order to allocate the different expenditure categories. All the forecasting models are put together in one equation which expresses the future expenditure substituted by fixed and mobile broadband, but which is still dependent on identifying the fraction. Detailed results are given in Deliverable D3.5.

Multilink techniques

MARCH investigated and developed multilink techniques and included these in the architecture. The techniques are either considered core network solutions or access network solutions.

The initial MARCH work reviewed relevant standardisation activities in ETSI, IEEE, IETF, and ITU. The project looked in particular to 3GPP and suggested split and merge of IP flows (sub-flows) as an additional feature to include in future releases. MARCH also worked out initial innovative combination of broadcast and communication networks with discussion of scenarios and architectures. However, in the second half of the project the focus had to be more limited and the focus was basically on wireless communication networks only.

MARCH provided an architectural framework helping structuring the ideas on multilink techniques and identifying the key areas to investigate. Furthermore, the background is the scenarios and business cases suggested.

The considered benefits from multilink techniques over single link access are

  • increased throughput,

  • increased reliability,

  • improved service continuity, and

  • higher network resources utilisation.

Either simple load balancing or adaptive including the more sophisticated splitting and merging methods can be used to achieve the benefits.

The reference end-to-end network architecture presented in Deliverable D4.2/D5.1 describes the multilink network elements with management decisions and control functions. It distinguishes between two different options deploying network level (IP and above) multilink techniques and radio access network techniques.

The network level techniques make use of single or multiple independent service data flows. The multilink gateway that resides in the core side is applied in the case a single multilink service data flow is being split and merged. The multilink adaptor on the subscriber side is matching the multilink getaway as it also is capable of splitting and merging of end user media streams.

Part of the architectural description focuses on how the general multilink techniques can be used in taking into account the emerging IP flow mobility techniques that is being standardised by 3GPP.

A simple way of operating the splitting of the media stream can be based on just a load balancing function, or a more intelligent splitting and merging set of techniques. MARCH focuses on the later, which can include techniques such as

  • adaptive video coding,

  • adaptive scheduling,

  • forward/backward error checking and correction, and

  • monitoring.

The radio access network techniques show how single link over radio access network’s performance can be improved by the following factors:

  • Service capacity, by selecting an optimal link for each user and using multiple input multiple output technology,

  • service reliability by using redundant links, and

  • increased quality of service by load balancing and user based parameters.

A number of multilink techniques have been investigated to reach these achievements, in particular:

  • Content adaption

  • Bandwidth aggregation

  • Service constraints optimisation

  • Forward error correction over multiple connections

  • Reliability and robustness/Reliability and security

  • Inter-system network load balancing with QoS awareness

  • Radio resource management

  • Multiple antenna techniques

Multilink network architecture

The MARCH multilink network architecture deals with operation of multiple networks for the same IP services, or IP flow, by introducing a multipath gateway. It also handles the case with off-loading of complete IP flows for a service by a mobile network based multilink gateway. The earlier Deliverable D4.2/D5.1 provided several options for an initial end-to-end multilink architecture and describes multilink techniques that are needed. This deliverable details the actual network architecture for solutions addressed in the scenarios (WP2), business (WP3) and demonstrations activities (WP6). The aim is to envisage multilink techniques in an end-to-end solution realisation and point to specific interfaces that are needed, whether it is a matter of specific usage or potential extension of existing ones or new interfaces.

MARCH focuses on solutions and opportunities around the following main topics

  • Multilink services for end customers enabling higher quality content streaming as well as live event distribution by using split and merge of single service data flow. These services are considered in different contexts involving a multipath gateway on the Internet (in the “cloud”) potentially provided by a 3rd party multilink service provider as an over-the-top approach with no direct impact on the mobile network operator or with some information provided by the mobile network operator or third party.

  • The mobile network operator focuses on use of multilink techniques and services for seamless Wi-Fi off-loading, improved network load balancing, and new services with improved user experience. This takes into account cases beyond the split and merge technique and includes also network triggered decisions and actions, as well as real-time network information for the purpose of improving the multilink adaptor performance and decisions.

  • Multilink features supporting emergency fleet support or disaster recovery solutions.

  • Network discovery and selection functions, considering various access networks discovery as well as network services discovery and selection topics, in various contexts of interest.

  • Multi-provider operations and business interaction support enabled by emerging business to business solutions for the support of multilink approaches. Examples are mobile network operator partnering with a hot spot network operator, automated roaming business and operations support, mobile network operators automated business and operations support for 3rd parties, and mobile network partnering with regional or local telecom operator.

MARCH has identified four main multilink connectivity modes: Service flow split/merge, i.e., the multipath gateway based mode, multiple TCP connections, IP flow mobility, and multiple access packed data network connectivity. All have an impact on what multilink techniques and functions that are needed in the various network elements. The modes can be used in combination. MARCH suggests a reference architecture providing the basis for the context and scope on specific multilink architectures. The architecture is aimed at short term multilink deployment, but is sufficient general for more future looking deployment as well. The key multilink techniques are positioned at the IP layer and above including a multilink gateway for the case where a service flow is split and merged for transmission over several physical accesses. In the user terminal there is a multilink adaptor, while the multilink gateway is located in the (core) network.

The benefit from deploying multilink access compared to individual link accesses is analysed through simulations and an analytical approach. Significant multilink gain, i.e., accommodated traffic using multilink technology compared the possible total traffic from using individual links without any cooperation, is possible depending on the load the request of communications links.

MARCH multilink architecture for mobile and fixed hot spot introduces a middleware to handle the packet flow over the various links. The mobility is kept through the mobile network access. Current smart phones and dongles are not yet able to handle multilink capabilities satisfactory, such that MARCH solutions should be considered for next version or generation of such terminals.

Multipath gateway can handle the split and merge of IP flows for technology neutral deployment. Several architectural options and issues are explored including multilink adaptations in the user equipment and in the gateway, in particular, multilink middleware and extensions of the socket API, utilising real-time local link layer performance information, and enabling feedback from the user equipment to the gateway for dynamically adapting its behaviour. One key issue is to explore whether real-time observations by the terminal can be conveyed back to the gateway for improved performance.

The MARCH network architecture control and management functions are described with a focus on user terminal with architecture for a generalised deployment of multilink. But the results are not limited to the split and merge case.

The radio access network architecture elements for multiple networks and access technologies are described including radio resource management, admission control, and based load balancing including quality of service considerations.

Mobile network offloading using Wi-Fi is considered with MARCH multilink architecture. The 3GPP architecture is introduced such as packet core gateway and dynamic policy-based service provisioning, to indicate where MARCH multilink capabilities can be included.

MARCH provides finally more detailed architectures for a mobile network operator partnering with a hotspot network operator, a mobile network operator partnering with a small independent Wi-Fi hotspot operator, and roaming with multilink and local breakout.


Part of the project results were demonstrated early, even before the work package demonstration had formally started. A fairly large number of demonstrations and trials have been given by project partners to show the feasibility of multilink techniques, such as

  • for a reporter uploading high quality video over narrow band mobile links,

  • resilient services that survives that one of the access links disappear and reappear,

  • robust multilink service for continued 24/7 video transmission in a difficult physical environment, and

  • radio resource management at core and radio access level.

Exploitation and dissemination

MARCH provided very early preliminary exploitation and dissemination plans, and has issued a separate project internal deliverable detailing the partners’ possible future exploitation of the results. This has been followed up with a deliverable at the end, D7.6, where each partner express views on exploitation plans.

A project web-site was early established for publication and the results as they have been achieved.

MARCH research has resulted in a number of articles in journals and conference proceedings, a number of presentations conferences and workshops, and public demonstrations.

All MARCH deliverables are public domain, except D7.6.

Expected impact

The impact from the results is positive for all type of partners in MARCH. Although estimate the total figure

Network operator

MARCH result contribute to network operators understanding of the broadband market end hence allow them to more precisely target the roll-out of service and network.

The result on business transitions between traditional ways of officering media, content, and telecommunication services enable operators to adjust the strategy for existing and new services.

Many operators have invested heavily in network infrastructure. The ML results on network efficient utilisation allow them to get more back form the investments and more carefully or softly enter into the next steps, such as roll out of LTE-networks.

Vendor industry

New products are needed for both the terminals and the network. It will basically be software products, but can include hardware products for new releases, for example from 3GPP, including ML technology such as split and merge of single IP flows.

Results on RRM is very suitable for WiMAX networks.

Telecommunication system developer

Advanced network services, such as for emergency situations or education purposes, are possible allowing system houses to develop new business. Also new network capacity providers are possible providing QoE enhanced services on top of exiting networks, for example in connection with major live events, e.g., bicycle race in a region or country. It is possible to set up new business with small investments.

One partner in MARCH, LiveU, has increased its business significantly in the period participating in MARCH.

Another partner, Lividi , was actually stabled at the beginning of the project and is now establishing business for real time video services that includes ML solutions.


The new knowledge allows academia to engage new doctor and master degree students, and develop mew or update existing courses with first hand knowledge on ML networking.

Abstract .

This report summarises some key results from all other MARCH deliverables. It is a selection covering multilink scenarios and economics, multilink techniques, multilink network architecture, multilink demonstrated, and some comments on expected impact. An annex is included with lists of all dissemination such as documents and papers produced.

Table of contents

  1   2   3   4   5   6   7   8   9   ...   16

База данных защищена авторским правом ©shkola.of.by 2016
звярнуцца да адміністрацыі

    Галоўная старонка