Liquid-State Drive: A Case for DNA Block Device for Enormous Data

This conference talk from FAST '25 explores the development of a DNA-based block storage device called LIQUID-STATE DRIVE for handling enormous data storage needs. Learn how researchers from Shanghai Jiao Tong University address the challenges of integrating DNA storage into silicon-based computer systems, focusing on petabyte-scale metadata management and reducing high DNA access costs. Discover their innovative dual-layer translation table approach that leverages SSDs to decrease metadata updating costs, along with symbiotic metadata and delayed invalidation techniques that significantly reduce garbage collection and block updating expenses. The presentation demonstrates impressive performance improvements in both microbenchmarks and real-world scenarios, with write cost reductions of up to seven orders of magnitude and read cost reductions of up to 6,206×. Gain insights into how this research bridges the gap between information technology and biotechnology, potentially accelerating the advancement of DNA storage for exabyte-scale data management.