Abstract
This disclosure describes techniques to virtualize input-output (I/O) on partitioned cloud servers in a manner that decouples the partitioning of compute resources (e.g., CPU cores, memory, etc.) from I/O, such that multiple compute partitions can share a single, securely multiplexed I/O infrastructure. Per the techniques, some read/write operations need not go directly to the corresponding hardware but may be handled by an intermediary layer that virtualizes the shared hardware based on the request and the partition from which the request originates. The intermediary layer supplies a response to the request, including the result of the read operation. The described techniques for partitioning I/O devices in a cloud-server environment complement techniques that divide a multi-core cloud server into multiple partitions, each of which runs a separate bare-metal OS or hypervisor potentially for distinct cloud customers. The techniques apply to commodity cloud server systems on chips (SoCs) using standard I/O interconnects such as PCIe, without the need for specialized devices.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
NA, "Input-Output Architecture for a Partitioned Cloud Server", Technical Disclosure Commons, (January 30, 2025)
https://www.tdcommons.org/dpubs_series/7782