We consider a distributed storage system employing some existing regenerate codes where the storage nodes are scattered in an optical wireless network. The data collector (DC) connects to the storage nodes via orthogonal channels and downloads data symbols from these nodes. In the existing data reconstruction schemes for distributed storage systems, the data collector downloads all symbols from a subset of the storage nodes. Such a full downloading approach becomes inefficient in wireless networks since due to fading, the wireless channels may not offer sufficient bandwidths for full downloading. Moreover, full downloading is also less power efficient than partial downloading. Given a coding scheme employed by the wireless distributed storage system, we propose a partial downloading scheme that allows downloading a portion of the symbols from any storage node. We formulate a cross-layer wireless resource allocation problem for data reconstruction in distributed storage systems employing such partial downloading. To derive the fundamental properties of partial downloading as well as to reduce the complexity of wireless resource allocation, we derive necessary and sufficient conditions for data reconstructability for partial downloading, in terms of the numbers of downloaded symbols from the storage nodes.We also propose channel and power allocation schemes for partial downloading in wireless distributed storage systems.
The purpose of distributed storage is to store a data file in a distributed manner where the individual storage nodes may be unreliable. This has attracted significant research interests in both communication and computer science fields. The original data file is firstly encoded into multiple coded symbols,which are stored into various storage nodes. Note that, encoded by advanced coding schemes, the original data can be reconstructed if the number of collected data symbols is no less than the number of original data symbols.
Recently, two data encoding schemes for distributed data storage have been proposed, based on rateless coding and network coding [1, 2]. A criterion for distributed data storage is the transmission bandwidth for the reconstruction of the original data file and the repair of a failed storage node. For data reconstruction, a data collector (DC) downloads the symbols in some storage nodes to reconstruct the data. For node regeneration, assuming that a storage node has failed, a new storage node downloads the symbols from some other storage nodes to regenerate the symbols in the failed node.