Systems Software for Fast Inter-Machine Page Faults

Cloud computing abstracts the underlying hard-ware details from the user. As long as the customer Service Level Agreements (SLA) are satisfied, cloud providers and operators are free to make infrastructural decisions to optimize business objectives, such as operational efficiency of cloud data centers. By adopting a holistic view of the data center and treating it as a single system, a cloud provider can migrate application compo-nents and virtual machines within the system according to policies such as load balancing and power consumption. We contribute to this vision by removing architectural barriers for workload migration and reducing the downtime of migrating processes. We combine the post-copy approach to workload migration with a novel specialized low latency interconnect for handling resulting remote page faults. In this work, we introduce a cross-architecture workload migration system, specify the requirements towards the specialized interconnect, discuss design trade-offs and issues, and present our proposed SW-HW co-design.

@article{Nider2018,
  title     = {Systems Software for Fast Inter-Machine Page Faults},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {9},
  number    = {1},
  year      = {2018},
  publisher = {The Science and Information Organization},
  author    = {Joel Nider and Mike Rapoport and Yiftach Binyamini},
  doi       = {10.14569/SpecialIssue.2018.0901102},
  url       = {https://doi.org/10.14569/SpecialIssue.2018.0901102}
}