With the increasing number of users in blockchain-based cryptocurrencies, the public has raised the demand for transaction throughput, and many protocols are designed to improve the throughput following the Nakamoto consensus. Although astonishing progress has been made in the on-chain throughput improvement, high throughput makes the blockchains suffer from the increasing blockchain size, hard forks, and possible attacks. In this work, we propose a quantitative model to describe and analyze the Nakamoto consensus. We then design a robust scheme named Ursa to reduce storage requirements and to reduce the forks by automatically adjusting block size according to users' needs.