Innovative Bloom Filter‐Based Multilevel Block Classification: A Practical Approach to Enhancing SSD Performance and Endurance

ABSTRACT In modern solid‐state drives (SSDs), managing hot and cold data is key to improving performance and extending lifespan. However, previous Bloom filter‐based methods were often complex and lacked solid empirical validation under high‐performance conditions. To address this, we propose a more efficient multilevel Bloom filter classification strategy tailored to SSDs. Our approach optimizes classification accuracy while reducing computational and storage overhead through careful parameter tuning. By utilizing SSD block‐level parallelism to group similar data access patterns, we enhance garbage collection efficiency and extend block life. Unlike previous studies relying on simulations, we validate our method on real SSD hardware. Experimental results show that our strategy improves SSD performance and endurance, offering valuable insights for future firmware optimization.