Reasearch Of The Memory Usage Optimization On The Stacked Depth Robust Grpah Algorithm

The distributed storage network based on Peer to Peer(P2P)is an important content carrier and core component of Web3.0.In order to combat the Sybil attack introduced by P2 P anonymity,distributed storage networks often use the Stacked Depth Robust Graph(SDR)algorithm to slow down the conversion of data copies.However,the intermediate data will occupy a large amount of memory,which becomes the main obstacle for idle devices to join the distributed storage network and parallel execution of multiple conversion processes.Currently,within the scope of our research,there is no solution to this problem.Through analysis of the structure and the execution process of SDR,the characteristics of fixed node dependencies and progressive generation of nodes are found.Based on the above two characteristics,a new algorithm is proposed,named SC-SDR.Specifically,for the fixed dependency of nodes,SC-SDR stores the intermediate data in the storage sequentially according to the usage sequence,thus greatly reducing the memory overhead.And the intermediate data accessed sequentially can maximize the use of storage bandwidth,thereby maintaining high algorithm performance.In addition.For the progressive generation of nodes,SC-SDR adopts a progressive memory allocation mechanism,which reduces the total amount of memory space occupied by multiple parallel execution SDR algorithm.Based on the above two memory usage optimization techniques,the final memory usage optimization algorithm SC-SDR is proposed.SC-SDR can greatly reduce the memory space overhead while maintaining high algorithm performance.The prototype of SC-SDR is realized and a lot of testing work based on the prototype is completed.The experimental results show that when the dedicated instruction set is not used,compared to the original SDR algorithm,SC-SDR can reduce the memory overhead of processing a 32 GB copy from 64 GB to 192 MB,that is,reduce the memory overhead by99.7%,and keep the algorithm execution time unchanged.When the dedicated instruction set is used,the execution time of the algorithm increases by about 49% while saving 98.4%of the memory overhead.