Dust Expiry: Clean the UTXO set from spam

Recent experiments focusing on the impact of syncing on ramdisk have shed light on the performance bottlenecks in blockchain node operations, particularly highlighting that when sufficient RAM is available, the efficiency of syncing processes tends to be constrained more by CPU and network limitations rather than IO costs. This observation underscores a critical inefficiency within the current design, where despite the availability of multiple CPU cores, the lack of parallelization leads to scenarios where typically only one core is fully utilized at any given time. This limitation becomes especially apparent when operating below the assumevalid height with default settings, pointing towards a systemic underutilization of available computational resources.

Furthermore, the introduction of Utreexo, aimed at reducing the RAM needed for fast validation, inadvertently does not contribute significantly toward enhancing the parallelization of these operations. Instead, its role is primarily confined to managing memory requirements more efficiently. The observations also reveal that the synchronization process leans heavily on CPU resources because the UTXO (Unspent Transaction Output) set remains within the bounds of available, unused RAM, making larger database cache (dbcache) settings redundant for nodes. This scenario is partly due to opportunistic write-through memory caching mechanisms employed by operating systems, which manage to keep the sync process CPU-bound as long as the UTXO set can be accommodated within the system's unused RAM.

These insights underscore a pressing need for reevaluating the architectural approaches to blockchain node synchronization, emphasizing the importance of developing more sophisticated parallelization techniques. By addressing these challenges, there is potential to significantly improve the efficiency and speed of blockchain operations, thereby enhancing overall system performance in environments where hardware resources are not a limiting factor.