Read disturb is a circuit-level noise in SSDs, which may corrupt existing data in SSD blocks, and then results in high read error rate and longer read latency. This paper proposes schemes of write scheduling and data reallocation, by taking account of read disturb. We first construct a model to estimate the block read error rate caused by read disturb, by referring the factors of block’s P/E cycle and the accumulated read count to the block. Then, the data being intensively read are flushed to the block having a small read error rate. Moreover, we introduce a data reallocation mechanism, which is completed by read reclaim, for balancing read accesses in all blocks. Thus, the total read errors introduced by read disturb can be cut down. Through a series of emulation tests based on several realistic disk traces, we demonstrate that the proposed mechanism can yield attractive performance improvements on the metrics of read latency and read error rate.