E2E-Clouds is a five-year research project financed by the Swedish Foundation for Strategic Research. The project is carried out by researchers at KTH - Royal Institute of Technology and the Swedish Institute of Computer Science (SICS Swedish ICT AB), Stockholm, Sweden. The goal of the project is to develop an End-to-End information-centric Cloud (E2E-Cloud) for data intensive services and applications.
The E2E-Clouds is a distributed and federated cloud infrastructure that meets the challenge of scale by aggregating, provisioning and managing computational, storage and networking resources from multiple centers and providers. Like some current data-center clouds it manages computation and storage in an integrated fashion for efficiency, but adds wide-scale distribution.
Some of the challenges addressed can be seen as overcoming the limitations of current clouds, many of which are far more important in a distributed cloud than the single data-center cloud. These limitations include:
- The promise of computing as a utility is not yet realized. Consumers should be able to easily move between and combine services from different producers. The infrastructure should be open to producers as well as consumers, enabling organizations to sell and buy resources depending on current needs. Contrast this with current clouds that are centrally managed and owned by few cloud providers without any interoperability among them. We assert that clouds need to be distributed, geographically and organizationally, aggregating geographically dispersed resources from many providers, ranging from large data-centers to smaller clusters (micro-clouds).
- The promise of elasticity is only partially realized. Adding new resources in terms of virtual machines still takes a considerable amount of time (i.e. minutes), making them unsuitable for the more demanding bursty applications, e.g., real-time media and entertainment applications.
- The resource allocation mechanisms are coarse-grained, leading to inefficiencies when different programming frameworks with different requirements run in the same cloud.
- Bandwidth is not treated as a managed resource, making it difficult to test and predict the behavior of distributed applications running in the cloud, even in the single data-center cloud.
- Existing software tools to support data-intensive and media-intensive applications lack functionality and are difficult to use.
- The lack of security provisions for protecting users from the provider preventing the adoption of cloud computing for mission-critical applications.