The semantic web is an extension of the web where information has a precise meaning. Thousands of linked datasets are available on the web. Important problems concerning quality, deep web access and availability still unsolved. For data quality, we propose to transform the web of data into a read/write web of data. A data consumer will able to correct an error. Allowing consumers to write the semantic web poses the problem of data consistency. We define synchronization algorithms for RDF data model. To access to the deep web, we propose a mediator approach allowing to combine semantic data and deep web data. The problem is to improve the performance of queries in the presence of a large number of data sources. Finally, to ensure the availability, we propose a replication model for the web of data. The problem is to optimize federated SPARQL queries in the presence of replicas selected at queries execution time.