VirtIO-FS as the shared file-system for transferring data between guest VMs to/from a host server is showing some promising performance results with pending code for allowing synchronous queues rather than always being asynchronous. Sergio López has been working on supporting synchronous queues for VirtIO-FS. VirtIO typically relies on asynchronous behavior but often times for I/O transfers you are specifically waiting on that to complete. Thus he pursued experimental synchronous queues for VirtIO-FS so the host thread running the vCPU process can complete the request before returning to the guest. The synchronous queues is showing massive performance potential with the time to build libkrunfw being used as a test, which dropped from 7 minutes and 43 seconds to just 5 minutes and 34 seconds in using synchronous queues. The sync queues also showed very significant improvements in random I/O performance, lower file deletion latency, and more. More details on this very promising work for VirtIO-FS via this blog post. There is a pending experimental pull request for libkrun adding the optional VIRTIO_FS_F_SYNC support. Very promising results so far for this implementation.
Synchronous Queues For VirtIO-FS Show Huge Performance Potential
Full Article
Original Source
Read the full article at Phoronix →KhanList aggregates and links to publicly available news content. We do not host full articles from third-party sources. Always verify important information with original sources.