The hashing method is widely used as the index structure, which can be stored in NVM to improve the application performance. However, existing hashing methods may cause high extra write amplification to NVM and bring high additional storage overhead on NVM while providing low request performance. To solve these problems, we have proposed a dynamic hashing method calledHierarchical Hashing, whose basic idea is to leverage a novel hash collision resolution mechanism that can dynamically expand the size of the hash table.Hierarchical Hashingcan incur no extra write amplification to NVM when resolving hash collisions. Additionally, it can directly address all cells when resizing the hash table, thereby avoiding the additional storage overhead caused by non-addressable linked lists. Furthermore, the request performance can be improved as all cells of the hash table are addressable when resizing to resolve hash collisions. The experimental results demonstrate thatHierarchical Hashingbrings no extra write amplification to NVM and achieves nearly 90% space utilization and high request performance while providing 99% memory utilization, compared with existing representative hashing methods.

Benefits from the simplicity, fastness and accuracy, the LogP model family is widely used to predict the parallel application communication performance, especial for the large-scale parallel application prediction or online prediction. However, this type of methods usually lacks consideration of modeling the network contention effect. This hinders their performance in predicting some real-world parallel applications. We thus propose a new parallel computation model via extending the LogGOPS mode with a parameter C. The additional parameter C is the additional time overhead caused by the network contention and it is predicted by a designed BP neural network. The experimental results show that LogGOPSC is more accurate than LogGOPS when there occur network contentions. Compared to the LogGOPS model, LogGOPSC gains a 93.25% average accuracy improvement for predicting 8mb point-to-point message passing. Furthermore, the average error of predicting two communication patterns is as low as 14.50% on the TianHe-2 HPC system.