Blockchain empowers a decentralized economy by enabling distributed trust in a peer-to-peer network. However, surprisingly, a widely accepted definition or measurement of decentralization is still lacking. We explore a systematization of knowledge (SoK) on blockchain decentralization by reviewing existing studies on various aspects of blockchain decentralization. First, we establish a taxonomy for analyzing blockchain decentralization in the five facets of consensus, network, governance, wealth, and transaction. We find a lack of research on the transaction aspects that closely characterize user behavior. Second, we apply Shannon entropy in information theory to propose a decentralization index for blockchain transactions. We show that our index intuitively measures levels of decentralization in peer-to-peer transactions by simulating blockchain token transfers. Third, we apply our index to empirically analyze the dynamics of DeFi token transfers. Intertemporally, we observe that levels of decentralization converge regardless of the initial levels of decentralization. Comparison of DeFi applications shows that exchange and lending are more decentralized than payment and derivatives. We also discover that a greater return of ether, the native coin of the Ethereum blockchain, predicts a greater decentralization level in stablecoin transfer that includes ether as collateral. Finally, we develop future research directions to explore the interactions between different facets of blockchain decentralization, the design of blockchain mechanisms that achieve sustainable decentralization, and the interplay of decentralization levels and economic factors.