Data deduplication schemes have been widely used in cloud storage systems to save storage space by eliminating duplicate outsourced data. However, the reduction of redundancy leads to decreased data availability and unbalanced data distribution. This paper proposes a blockchain-based decentralized storage system with reliable deduplication and storage balance strategy to provide reliability for deduplicated outsourced data. Encrypted data is split into chunks by a ramp secret sharing scheme, and it is distributed to multiple independent cloud servers. States of the chunks are recorded on the tamper-proofed blockchain and they can be used to recover the raw data or support the verification of user identity. To balance the distribution of data among storage servers, a heuristic matching algorithm is designed to efficiently allocate the available storage space. The allocation services are published by autonomous smart contracts and other participants gain rewards by giving the best matching results of data chunks and storage servers. Formulation analysis demonstrates the correctness of the proposed scheme in terms of data consistency, integrity, and reliability. Experimental results show that the proposed scheme preserves the confidentiality of outsourced data with acceptable computational consumption.