I/O for wide-area systems

Overview: 

logoThe wide-area computing (Grid, Cloud, ...) is becoming one of the key computing paradigms for the near future. In order to make it widely usable, many data management problems need to be addressed. Some of them are related to data movement, data accessing, and usage semantics.

Objectives: 
  • Replica creation and deletion. Most system currently create replicas on a on-demand basis. This approach reduces the performance benefit that can be obtained by replication. Our proposal is to research on smart mechanisms that decide when replicas need to be created before the replicas are really needed. In addition, we will also use similar mechanisms to decide when a replica is of no use anymore before capacity problems appear.
  • Grid file systems based on object storage devices. We will integrate the above mentioned mechanisms, plus basic replica management, in OSD-based file systems (XtreemFS) where management is fully distributed. This lack of centralized information will be a challenge in order to implement the above-described mechanisms.
  • Efficient access to read/write data from the cloud. Cloud computing and cloud storage are becoming a well used paradigm to run applications. On the other hand, the access of shared data that is read and written concurrently is still an open problem that significantly limits the usability of such environments. Our group is devoting some effort to improve this kind of access patterns in cloud environments.
Additional Information: 

Contact: toni.cortes  bsc.es

European Projects