RELATED WORK

Commercial Solutions: Technologies from Qualcomm [29] and Interdigital [30] provide WiFi offloading with intelligence mainly at the mobile devices. They claim to manage user flows across WiFi and 3G/LTE networks based on throughput and delay measurements. They rely on integration with the I-WLAN architecture to provide seamless connectivity. Although not widely deployed, the existence of such solutions indicates the importance of managed WiFi offloading. Since these technologies incorporate context in their solution and provide management across 3rd party WLANs that are not managed by the operator, they can be used as complementary solutions to ATOM.

商业解决方案：

来自高通[29]和Interdigital[30]的技术主要在移动设备端提供智能化的WiFi分流功能。它们声称能基于吞吐量和延迟测量来管理跨WiFi和3G/LTE网络的用户流，并依赖与I-WLAN架构的集成来提供无缝连接。尽管尚未广泛部署，这类解决方案的存在表明了受管理的WiFi分流的重要性。由于这些技术在其解决方案中融入了上下文感知，并提供了对非运营商管理的第三方WLAN的管理，因此它们可以作为ATOM的补充解决方案。

Client-side Optimizations: Recent studies [7, 8, 9, 10, 31] have proposed distributed algorithms for interface selection. Most of these solutions either require additional signaling from the network or leverage P2P to disseminate the current network. As shown in our evaluations, even with network load information, such solutions are not as efficient as a centralized solution such as ATOM. Some studies also argued the feasibility of using public WiFi for offloading 3G traffic [11, 12]. However, the scope is limited to delay tolerant traffic and may require changes to the applications to support the framework. There is also a recent study [32] that proposes seamless switching of user traffic but it does not provide any intelligence for optimal interface selection.

客户端优化：

近期的研究[7, 8, 9, 10, 31]提出了用于接口选择的分布式算法。这些解决方案 大多需要来自网络的额外信令，或利用P2P技术来传播当前网络状况。正如我们的评估所示，即使拥有网络负载信息，这类解决方案的效率也不如像ATOM这样的中心化解决方案。一些研究也探讨了使用公共WiFi分流3G流量的可行性[11, 12]。然而，其范围仅限于时延容忍型流量，并且可能需要修改应用程序以支持其框架。近期还有一项研究[32]提出了用户流量的无缝切换方案，但它并未提供用于优化接口选择的任何智能机制。

Network Solutions: Several studies have proposed network-driven algorithms for interface selection or user association to optimally balance the load across heterogeneous networks. A few of these works [6, 33, 34, 35] use utility-based optimization to maximize the throughput obtained by the users. However, these works (i) assume idealized settings with little or no consideration of practical constraints (ii) are tightly integrated with the basestation schedulers hindering their deployability and (iii) do not provide an end-to-end solution that dynamically selects interfaces adapting to changes in the mobile network. ATOM on the other hand is a comprehensive solution for traffic management for heterogeneous WiFi-LTE networks that is adaptive, light-weight, scalable and deployable in today’s mobile networks.

网络解决方案：

若干研究已提出网络驱动的接口选择或用户关联算法，以在异构网络间优化负载均衡。其中一些工作[6, 33, 34, 35]使用基于效用的优化方法来最大化用户获得的吞吐量。然而，这些工作存在以下问题：(i) 假设了理想化的设置，很少或没有考虑实际约束；(ii) 与基站调度器紧密集成，妨碍了其可部署性；以及 (iii) 未提供能够适应移动网络变化、动态选择接口的端到端解决方案。与此不同，ATOM是针对异构WiFi-LTE网络的全面流量管理解决方案，它具有自适应性、轻量级、可扩展性，并且可以在当今的移动网络中部署。