Performance Modeling And Analysis Of Hybrid Page Cache Systems

In recent years,there is an increasing demand of large-capacity page cache systems so to perform large scale data analytics and high performance computing.Since DRAM memory is expensive,some researchers are suggesting to combine other memory tech-nology such as non-volatile memory(NVM)technology to build a large-capacity hybrid page cache system.However,whether the NVM technology can improve the perfor-mance of the whole cache system(either economically or technically)remains an open question.To answer this question,it is important to consider how to design a hybrid page cache system efficiently incorporating different performance characteristics and price ratios of hybrid memory devices.This thesis presents an analytical model of a hybrid page cache system so to un-derstand the diverse design space and performance impact of a hybrid cache system.We consider(1)various architectural choices,that is," Flat Architecture " and " Lay-ered Architecture",(2)design strategies,and(3)configuration of different memory devices.We first describe the workload model,then characterize the dynamics of data pages in hybrid cache by using the Markov model,and finally derive the cache content distribution in steady state.Further,we conduct mean-field analysis to approximate the cache content distribution so as to make the computation more efficient.After that,we define a latency-based performance metric based on the cache content distribution so as to quantify the overall cache performance.Through numerical analysis based on the model,we provide some guidelines on how to design hybrid page cache to reach a good trade-off between high system through-put(I/O per sec or IOPS)and fast cache reactivity which is defined by the time to fill the cache.We pick PCM as an example for NVM and conduct numerical analysis.Our analysis indicates that incorporating PCM in a page cache system significantly improves the system performance,and it also shows larger benefit to allocate more PCM in page cache in some cases.Besides,for the current setting of performance-price ratio of PCM," flat architecture " offers as a better choice,but " layered architecture " outperforms if PCM write performance can be significantly improved in the future.