The Middleware for Edge Clouds & Cloudlets (MECC) workshop aims to bring together researchers and practitioners in an effort to achieve a better integration between the different tiers on modern cloud computing platforms. There is a growing trend of interactive and resource-intensive (e.g., compute, storage, need for big data) applications on mobile devices. Currently, many of such applications are supported by remote resources on infrastructural clouds. However, to efficiently access and mange such resources when faced with limited and/or intermittent connectivity is challenging. An increasingly viable alternative is to harvest resources present on nearby mobile devices and/or cloudlets. Today, there is also increasing demand for middleware that offers higher level abstractions without hampering expressiveness and perfor- mance. However, most of current distributed systems are designed for the datacenter, and their assumptions, such as that nodes use fast wired interconnects, no longer hold in edge environments. In particular, edge clouds, such as those made up of only mobile devices at the edge, use unreliable wireless links. These unreliable links directly translate into unavailability and churn. Simultaneously, since mobile devices have limited energy resources, heavyweight distributed algorithms, such as coordination using a leader-based consensus protocol, are impractical. As an effort to offload computation from mobile devices, cloudlets were originally envisioned as server-class hardware deployed in a neighborhood, office building or more generally, in close physical proximity to any scenario with a high density of users, such as at large public events. It is now transitioning to a more lightweight approach where the offloading is done through multiple techniques besides the use of virtual machines, as originally proposed, and where cloudlets can also offer connectivity support to crowd-sourced mobile devices, i.e., edge clouds. With this new trend in sight, there is a need to define the services that should be offered at each tier. For example, cloudlets can provide well-defined APIs to support multiple computation offloading methods. Furthermore, new modular and reconfigurable architectures have to be proposed in order to support a variety of deployment scenarios, such as edge clouds without cloudlet support, and scenarios with very limited access to infrastructural clouds.
|Publisher||Association for Computing Machinery, Inc|
|Number of pages||25|
|Publication status||Published - Sep 2016|