ABSTRACT:
With the growing importance of in-memory data processing, cloud service providers have launched large memory virtual machine services to accommodate memory-intensive workloads. Such large memory services utilizing low volume scaled-up machines are far less cost-proficient than scaled-out services consisting of high volume commodity servers. By exploiting memory utilization imbalance across over cloud nodes, disaggregated memory can scale up the memory limit with regards to a virtual machine in a cost-effective way. Disaggregated memory enables accessible memory in remote hubs to be utilized for the virtual machine requiring more memory than its locally available memory. It supports high with the quicker direct memory while fulfilling the memory limit request with the slower remote memory.
This project proposes another hypervisor-integrated disaggregated memory system for distributed computing. The hypervisor-integrated outline has a few new commitments in its disaggregated memory plan and usage. Initially, with the tight hypervisor coordination, it examines another page administration instrument and approaches tuned for disaggregated memory in virtualized systems. Second, it rebuilds the memory administration methods and relives the scalability concern for supporting huge virtual machines. Third, exploiting page access to records accessible to the hypervisor, it supports application-aware elastic book sizes for getting remote memory pages with various granularities. Contingent upon the degrees of a spatial region for various locales of memory in a virtual machine, the optimal block size for every memory area is dynamically selected. The experimental results with the usage coordinated to the KVM hypervisor, demonstrate that the disaggregated memory can give all things considered performance degradation compared with the ideal local memory just machine, despite the fact that the direct memory limit is only of the total memory footprint.
DOWNLOAD: Disaggregated Cloud Memory with Elastic Block Management