Summary: The research paper studies on of the important and common data structure called B+Tree. The consistency cost of the data structure is the topic of the research. The researchers have implemented NV-Tree on NVDIMM server and then evaluated its performance. The results suggest that the B+Tree performs better than most of latest consistent tree structures by a multiple of 12.
Strengths: The researchers have used a model that compares different other consistent data structures to reach to make accurate suggestions about the performance and consistency of B+Tree. NV-Tree has been used as the core data structure which reduces the workloads by more than 90%. NV-Tree and NV-Store are implemented and evaluated on a real NVDIMM instead of any simulator.
Weaknesses: the research has not considered the cost of L3 cache misuse.
Questions: What will be the effects of the overhead in the rebuilding of larger datasets?
Will the performance of NV-Tree be validated under skewed and TPC-C workload?
How will NV-Tree respond to a distributed environment?