Chapter 5
Conclusions
In the scope of beyond-3G heterogeneous environments, the goal of the Ambient Networks project is to realise a system that provides access to any network, including mobile personal networks, through instant establishment of inter-network agreements in accordance to user and/or operator preferences. Thanks to negotiations between the involved entities, each Ambient Network becomes aware of the surrounding environment and can take appropriate decisions for the different steps of the handover process, exploiting a variety of triggers in a multi-dimensional mobility management and HO management.
Our work has dug into HO management issues within the heterogeneous environment of Ambient Networks. In particular, concepts have been developed along with a realization example that has been specified and implemented.
In the first section of this chapter we describe the implemented solution. Mobile IPv6 [1], Fast Handovers for Mobile IPv6 [2] and HO Toolbox implementation [3] are compared. We highlight how a sharp distinction between macro and micro mobility could lose significance in future mobile environments, leading to our choice to design a new mobility protocol. Section 5.2 shows the objectives achieved, while the following section is dedicated to the description of possible future developments.
5.1. HO Toolbox implementation
The HO Toolbox implementation could be considered as a prototype of a new mobility protocol that demonstrates a number of AN concepts going beyond traditional mobility concepts. It uses concepts of both MIPv6 and Fast HO, trying to combine their benefits without expressly using their exact protocol descriptions, as available from the IETF, and without being based on existing implementations of them.
Common IETF literature is classically distinguished between macro (e.g. MIP) and micro (e.g. Fast HO) mobility protocols, even though Fast HO draft is designed to complement MIP, i.e. it is an extension to MIP.
Likely, future mobile environments aren’t characterised by a clear distinction between macro and micro mobility, as beyond-3G networks are not necessarily subdivided between macro and micro domains. Taking for instance Fast HO: future peer-to-peer scenarios may also include the case where the PAR (as defined by Fast HO) is also a CN (as defined by MIP), therefore creating some confusion regarding how e.g. FBU/BU messages (which have the same format) should be interpreted.
Moreover, macro-mobility typically refers to reachability and route optimisation issues in IP networks, while micro-mobility typically refers to local mobility solutions and inter-access router handovers. The Ambient Networks project in its first phase tries instead to look at mobility from a broader viewpoint, even if migration issues from current systems are also considered.
For these reasons, we decided not to distinguish between macro and micro mobility, and we realised one single mobility protocol that can handle both situations, still using concepts of both MIP and Fast HO.
Our implementation scenario considers four Ambient Networks (AN) and the following mapping: AN1 is a CN, AN2 is the MN, AN3 the PAR and AN4 the NAR. Flows are established between AN1 and AN2 through AN3, and handed over to a new route passing through AN4.
Each AN includes a minimal Ambient Control Space (ACS) that acts according to our mobility protocol, exchanging messages with the other ACSs. To be noted that this inter-ACS message exchange represents an example of ANI messaging.
In addition, it is possible to run end-user services between the different ANs. For realising this, typical applications were used for sending packets to an IP address that we will call Home Address (HoA) for simplicity (which could also be considered a primary CoA).
Session mobility was realised by using multi-homing: each service/flow is mapped to a single HoA of the same AN, so that we can implement handover and forwarding of a single service/flow independently. Handovers of single sessions may be required directly by the user or by user/network policies, possibly over multiple access network technologies, hops and administrative domains (e.g.: roaming).
The mobility protocol we designed is based on exchange of “AN Negotiation” messages called ANeg, providing a common framework for all ANI messages in our implementation. ANeg messages are built as ICMP messages within respective options.
ANeg messages inherit the characteristics of both RtPrSol and PrRtAdv [4]
of Fast HO, therefore they can be both soliciting or advertising messages. The nomenclature of typical Fast HO messages such as FBU, HI, HAck, FBack, FNA is reused, but the meaning of each message is largely extended, while the message itself is translated into an Ambient Netgotiation message in the sense that it carries ANeg options.
The FBU message is transleted into an “AN-FBU message”, extending its significance in the field of multi-access environments: it includes the HO
Toolbox Option indicating whether to continue deliver on the old link, or
NAR forwarding, or bicasting. The latter may be useful in particular cases, e.g. when it is crucial not to lose packets. An AN-FNA would then be needed to switch to the new link, or to stop bicasting on the other link. In this sense, the meaning of the FNA message was changed, not to indicate only the link attachment detection, but also a possible handover command. Instead, if we would like to emulate the case of the so called “Predictive Fast HO” of Fast HO [5], the handover command should be placed already in the FBU message. AN-HI and AN-HAck messages are extended with the Composition
Option, a concept proper of the Ambient Networks project.
5.2. Objectives achieved
The HO Toolbox implementation and reference scenario are useful examples of the characterising ideas of the Ambient Networks project. The HO toolbox exemplifies how heterogeneity can be achieved, allowing to construct handovers flexibly based on actual needs and requirements within an Ambient Networks environment. The reference scenario shows a typical
setting for Ambient Networks communications, including e.g. an operator network (AN1) as well as a PAN (AN3 and AN2) and a local access provider (AN4). This section illustrates how HO Toolbox implementation satisfies the requirements of this reference scenario.
In mobile environments there could be necessity to prevent packet loss for some applications. The HO toolbox allows preventing packet loss thanks to commands/functions for resource reservation, buffering and delivering of packets. These commands are included in the Service, Bearer Service and
HO Toolbox options.
For other applications there could be the need to minimise the interruption time during handover, therefore bicasting could be the solution for this matter. As in the previous case, Service, Bearer Service and HO Toolbox options allow solving this problem.
In some cases instead, there could be opportunity or necessity to change the route(s) in order to improve QoS (in terms of packet loss, delay, jitter delay, etc.), to maintain QoS (e.g. when a link is going down) or to handover to a cheaper link. One more time, Service, Bearer Service and HO Toolbox options meet these opportunities/requirements.
In case of traditional handovers driven by a decrease in power strength, the Handover Toolbox allows obtaining handover latency reductions, given in this implementation that a valid IP address is formed before starting the effective handover procedure.
Other ANeg options are designed principally to meet requirements of composing Ambient Networks, but need to include information useful in the preliminary phases of handover such as detection of potential alternative routes, etc. So, Composition and Access Route options, also including the IP
created with the aim to allow discovery of required characteristics in other Ambient Networks and to compose with them.
Note that the Handover Toolbox implementation of the reference scenario encompasses almost all basic concepts proper of the Ambient Networks project that have a relation to mobility and handover. In other words our work represents an AN prototype that implements the HO toolbox.
Dynamic composition of networks is shown in the “service agreement” between AN1 and AN3, in the “PAN composition” between AN3 and AN2, in the service agreement between AN3 and AN4. Moreover handover preparation requires the same type of composition between AN2 and AN4 via AN3, where messages AN-HI and AN-Hack are exchanged.
Thanks to composition, each AN can use other ANs to obtain more user and network services exploiting different access technologies and realizing handovers in the Ambient Networks multi-dimensional environment.
Usage of multiple devices, even belonging to different users, is visible in the communication between the four ANs involved: AN2 uses either AN3 or AN4 as the access point for a given service. The handover performed is also multi-hop because the download service is hijacked to AN2’s eth1 card via AN4, but it is also a service based handover, because it can be performed for single sessions only. Moreover, it is a multi-domain handoff as well, given that AN2 and AN3 belong to the same domain, but not AN2 and AN4: in fact during the AN-HI/AN-Hack negotiations AN3 performs an A/A/R for AN2 at AN4 leading to the composition between the two domains.
In this reference scenario we did not use different accesses technologies, but only used cables. This is due both because we focused on higher level messaging and because the nature of this implementation is inherently above IP. Future development steps will however include multiple wireless
accesses, as well as other lower-layer Ambient Networks implementations. Note that the usage of cabled accesses only is not a restriction, since e.g. the access between AN2 and AN4 could have been made WLAN easily, introducing the appropriate commands for IP level configuration.
The Ambient Networks project is currently in a conceptual phase which could result in a number of possible implementations at different levels e.g. above or below IP. The work of our implementation could be considered as a first step towards a clean redesign of IP mobility based on explicitly stated principles and requirements, in contrast to the idea of adding patchworks and extensions to existing protocols and architectures. A different implementation option for the HO Toolbox concept could be e.g. the one to realise the HO toolbox as a set of mobility protocols and extensions, to be selected depending on specific mobility needs. However, final decisions on the allowed Ambient Networks implementations will not be taken during the first phase of the project, while they are expected before 2008.
5.3. Future developments
Next steps will provide further development of code and extensions of the reference scenario, in line with typical Ambient Networks mobility scenarios.
A first step will be to substitute some cable links with wireless links, e.g. a traditional 2G/3G link between AN1 and AN3, Bluetooth between AN2 and AN3 and WLAN between AN2 and AN4. Still, this is not expected to have a strong impact on our type of implementation.
In order to demonstrate as much Ambient Networks concepts as possible, more flexibility will be introduced in the scenarios and more sub-cases will be generated, e.g. using the implemented buffering functions to avoid packet loss, or using bicasting to minimize the interruption time. According to this, HO management and the embedded HO toolbox, along with their capabilities in such scenarios will be demonstrated.
Future work may also develop around dynamic network composition scenarios. For example, during a urban journey on a bus, several mobile devices could use some wireless access points available along the route, belonging either to operators or other users (local access providers). Inside the bus a number of devices composes together in an Ambient Network, in such a way that only one user is selected to provide connection towards the outside due to specific reasons. This selected user would be a moving gateway for the users within the bus and may have several benefits, e.g.: gaining traffic bonuses or credits, or priority for obtaining services from others. This scenario could lead to reduce signaling with the exterior, and to share all available resources dynamically, also boosting to creation of new possible applications (e.g.: gaming, entertainment, teamwork, etc.) and new socio-economic scenarios.
5.4. Final remarks
The results of this thesis have been obtained into the Ambient Networks project through Siemens, and they already had an impact with respect to the developed HO concepts. Moreover, they will probably have a future impact on the ANI interface design, which is of fundamental importance for the
project. Additionally a future impact in determining possible implementation options is possible.
If the presented type of implementation will be accepted and pushed by the project, additional refinement of code and kernel space implementations will be required, as well as contributions to IETF may be generated.
However, Siemens will continue this work within Ambient Networks in this and the next phases. Goals include obtaining better network possibilities and performances, extended networking capabilities and services for the users, but above all to guarantee a continued growth in the mobile communication industry for the sake of manufacturers, operators and communities. Operators would benefit by selling new services; manufacturers would benefit by selling new Ambient-Networks-capable devices and access networks; the community will enjoy the improved networking capabilities and enhanced availability of services, with all the effects and business/social scenarios that such new networking means might generate.
