A model of distributed storage split data using algorithms, multidimensional parity, resistance to partial losses of the storage sites, is regarded as an alternative way of security that can replace the conventional multiple reservations and carrying costs of growth of the physical volume. In this model, the generated redundant data allows you to restore partial loss of the split parts. In this case, the recovery is performed at the expense of the parity files formed during the splitting procedure, in which the main action is the calculation of bitwise parity with summation modulo "two". With this additional operation between the source files, you can restore the corrupted file without distortion. A comparison with several models that use checksum codes for recovery shows that the estimated probability of losses due to storage failure in the case of multidimensional parity codes is significantly less than when using other split data storage options. Using the operation of adding the bits of undamaged files, you can completely restore corrupted files. According to the calculated data, even when theoretically non-recoverable losses are reached, it is always possible to find combinations of chain recovery. The complexity of alternative methods, including the iterative decoding method, makes a distributed split data storage system using multidimensional parity algorithms a more convenient error correction technology. Keywords— , , , , loss   recovery, .