In view of data intensive cloud computing applications

disk-based storage systems are difficult to meet their demand for performance and availability simultaneously.This paper puts forward a key-value distributed storage system named M-Cloud

which uses memory as main device and disk as auxiliary when providing storage services(read

write

backup and recovery)for big data.M-Cloud improves overall system performance by loading entire data in the memory of the cluster

and designs PBLH(partition-based linear hashing) algorithm to achieve load balancing and high scalability.M-Cloud also designs relative backup and fast crash recovery strategy in order to ensure high system reliability.The simulation results show that M-Cloud has high performance and availability

with further optimization

the system has the potential in real production environment usage

can provide users with high-quality storage services.