In this paper we analyze main use cases for sharing wireless access networks between multiple operators and service providers. Network sharing has been proposed as a metho l to lower roll-out costs for 3G operators in Europe, and is widely used in WLAN systems where local access providers offer wireless access to service providers. A similar structure also exist in cellular networks where Mobile Virtual Network Operators (MVNO) provide mobile services without having a mobile network of their own. The development points at a further fragmentation of wireless access networks into specialized service providers that connect to local service and access providers, possibly via an inter-connection provider serving with core network functionality. In this context, we propose a framework for how radio resources could be managed using Service Level Agreements (SLA) and analyze key differences between the SLA for different types of service and network providers.

An efficient combination of radio access technologies, integrated in a multi-access network, will be a key enabler in future provisioning of mobile data services. This paper addresses a scenario where an incumbent mobile network operator, with an existing 2G and 3G infrastructure, has deployed a dense WCDMA/HSPA macro-cell network in an urban area. With this high capacity network deployed, upgrading previous 2G and 3G systems with EDGE and HSPA respectively may seem obsolete. However, even though these systems may not support the intended data rates alone, they could, thanks to favorable propagation characteristics and the additional spectrum available, be useful as complements. Simulation results indicate that upgrading GPRS base stations with EDGE, or a sparse WCDMA macro cell layer with HSPA, mainly would benefit uplink transmission. For this case, the data rate that can be guaranteed with 95 % area availability (coverage) can be increased with approximately 40-100 %. In the downlink, though, the dense WCDMA/HSPA system alone supports user data rates of 500 kpbs for all relevant user densities. Thus, upgrading legacy infrastructure would be obsolete.

Radio resource management (RRM) across multiple, potentially competing, wireless networks has emerged as a salient feature for future generation system. Besides increased overhead, it will incur more complicated architectures and the question then becomes whether the gains of cooperative RRM (increased throughput, reliability, etc.) can justify these disadvantages. Herein we study potential gains that can be achieved by utilizing sophisticated algorithms in a scenario with two cooperating IEEE 802.11a networks, limited by either the wireless or wired link. Both best-effort (BE) and minimum bitrate (MBR) are treated and throughout the paper we use the rudimentary minimum path-loss (MPL) allocation as reference. Our results indicate that sophisticated access selection methods that besides path-loss, also account for AP load and potential constraints in the wired connection, can increase performance significantly. The maximum gain varies between 30-60% and 100-150% for BE and MBR traffic respectively and is typically obtained when there, on average, is one user per AP. We also studied the case where APs with constrained wired capacity connected (using the MPL criteria) to high-capacity APs in order to reroute traffic. Contrary to the approach in which advanced allocation principles is used, "loose" network integration is sufficient and even though it gave lower gains for BE data, the achievable rates for MBR data could be improved substantially.

An important question for future wireless networks is whether the prioritization between different accesses should be controlled by the networks or terminals. Herein we evaluate the performance of distributed access-selection algorithms where terminals are responsible for both AP selection and the necessary measurements. In particular, we focus on determining whether selfish distributed algorithms call perform as well as centralized ones (for comparison we include max-sum, max-min, proportional fair and minimum delay allocations). The study is conducted by time-dynamic simulations in a IEEE 802.11a network and its performance measures we use file transfer delay and supportable load at a maximum tolerable delay. Our results show that selfish algorithms can offer similar performance. both in terms of throughput and fairness, as the centralized schemes as long as they account for both path-loss and access point load. This is an important result and it suggests that terminal-controlled algorithms are Just as efficient as centralized schemes, which besides extensive measurements also require that AP exchange information, for improving the efficiency in WLAN networks. Compared with a minimum path-loss selection criteria, which is standard in the IEEE 802.11 family today, our distributed load-aware algorithm increases the maximum supportable load with more than 200 percent even after accounting for measurement time and estimation errors. With fast reselection during ongoing sessions the gains call be further increased with, typically, 20 percent.

For voice, an efficient radio resource management (RRM) essentially boils down to providing a predefined signal to interference ratio (SIR) at lowest cost possible and centralized schemes has, evidently, been an effective approach to address these problems. Delay-elastic data services, however, introduce both heterogeneous user requirements and possibilities for opportunistic RRM. One way, among others,to handle this would be to let autonomous trade-agents, acting on behalf of users, manage the radio resources, and this is our point of departure. We propose a market-based framework for decentralized RRM in environments populated by multiple, possibly heterogeneous, "access points" (APs), and the provided service for the users consists of file transfers. Resources (transmission time) are partitioned between users through a proportionally fair divisible auction. The problem at hand for the user (trade-agent), is then to determine how much resources it should purchase from the different APs in order to maximize its utility ("value for money"). Our results indicate that decentralized selfish bidding strategies are able to capitalize on temporary beneficial conditions and offer comparable performance with a centralized scheme (based on the 'mu C-rule') that requires knowledge about peak data-rates, queue lengths, and preferences for all users in the system.

Radio resource management (RRM) across operator boundaries is emerging as a salient feature for wireless systems beyond 3G. Until recently, research has been confined to solutions where cooperating networks enter explicit sharing agreements that define how responsibilities and revenues should be divided. An alternative would be to share the infrastructure implicitly by establishing an open wireless access market wherein networks not only compete for users on a long-term time-scale, but also on a much shorter time-base. This could be realized with an architecture where autonomous trade-agents, that reside in terminals and access points (APs), manage the resources through negotiations. In this paper we develop a framework for studying demand-responsive pricing in contexts where APs with overlapping coverage compete for users. Resources are partitioned through a proportional fair divisible auction and our aim is to establish if, and when, an open market for wireless access can be se sustained. Compared to a scenario where APs cooperate, our results show that, an open access market results in better services at lower price which in the prolonging also yields more satisfied customers. As an effect demand will increase and, from the perspective of the APs, act as a counterbalance to the reduced prices. Thus, the revenue earned by the AN will be comparable to the one in which obtained through AP cooperation and monopoly (cartel) pricing. Generally speaking, the difference between the cooperative and noncooperative RRM is small when the demand is concave and increases with the convexity of demand.

 Exploiting user-deployed networks as part of a public infrastructure has been proposed as a concept to radically lower the cost of provisioning access and services in urban environments. An important question is how dense a network with spotty coverage has to be in order to support interesting services. In this paper¹, we introduce a framework for evaluating the user-perceived performance for two service types, a time-critical news subscription service (involving streaming) and a more delay tolerant entertainment service ("web browsing"). The user terminals contain a memory cache and an "agent" that opportunistically downloads and stores relevant pieces of information as users walk close to the access points. Results show that above a critical access point density, that is still significantly lower than the one required for continuous coverage, the services start performing very well with little or no outdated information. In most studied scenarios, we are communication limited meaning that memory capacity is not a problem, but energy is.