Red Hat Bugzilla – Bug 633394
[6.1 FEAT] virtio-blk ioeventfd support
Last modified: 2013-01-09 18:08:05 EST
1. Feature Overview: Feature Id: [67241] a. Name of Feature: [6.1 FEAT] virtio-blk ioeventfd support b. Feature Description Virtio ioeventfd increases performance and scalability by decoupling guest code execution from virtio hardware emulation, reducing the amount of time stolen from the guest and allowing guest CPU and I/O to run in parallel. 2. Feature Details: Sponsor: LTC Open Virtualization Architectures: x86_64, Arch Specificity: purely common code Affects Kernel Modules: Field does not exist Delivery Mechanism: Direct from Community Category: other Request Type: Package - Update Version d. Upstream Acceptance: Field does not exist Sponsor Priority P3 f. Severity: normal IBM Confidential: No Code Contribution: IBM code g. Component Version Target:--- 3. Business Case Improved performance of virtio-blk and virtio-net devices in KVM VMs to better compete with other industry hypervisor solutions. 4. Primary contact at Red Hat: John Jarvis jjarvis@redhat.com 5. Primary contacts at Partner: Project Management Contact: Michael W. Wortman, wortman@us.ibm.com Technical contact(s): Stefan Hajnoczi, stefan.hajnoczi@uk.ibm.com
What is the component for this and is it upstream? Please provide more details about is being requested here.
------- Comment From stefan.hajnoczi@uk.ibm.com 2010-09-22 05:18 EDT------- The component is userspace QEMU. Patches will be sent for upstream soon, the code is written.
------- Comment From rsisk@us.ibm.com 2010-10-04 10:40 EDT------- Code Upstream Status: Not evaluated
------- Comment From stefan.hajnoczi@uk.ibm.com 2010-11-15 08:17 EDT------- Apologies, the status should be OPEN. I have submitted patches upstream but they have not been merged and I have not provided a backport yet.
This request was evaluated by Red Hat Product Management for inclusion in the current release of Red Hat Enterprise Linux. Because the affected component is not scheduled to be updated in the current release, Red Hat is unfortunately unable to address this request at this time. Red Hat invites you to ask your support representative to propose this request, if appropriate and relevant, in the next release of Red Hat Enterprise Linux. If you would like it considered as an exception in the current release, please ask your support representative.
This request was erroneously denied for the current release of Red Hat Enterprise Linux. The error has been fixed and this request has been re-proposed for the current release.
IBM is signed up to test and provide feedback.
This enhancement request was evaluated by the full Red Hat Enterprise Linux team for inclusion in a Red Hat Enterprise Linux minor release. As a result of this evaluation, Red Hat has tentatively approved inclusion of this feature in the next Red Hat Enterprise Linux Update minor release. While it is a goal to include this enhancement in the next minor release of Red Hat Enterprise Linux, the enhancement is not yet committed for inclusion in the next minor release pending the next phase of actual code integration and successful Red Hat and partner testing.
I have successfully tested RHEL6.1 qemu-kvm 0.12.1.2 2.148.el6 with libvirt 0.8.7 7.el6. ioeventfd is functioning as expected. Here are the tests I performed on a -smp 2 guest: 1. RHEL6 install to 100 GB virtio-blk raw file. 2. RHEL6 install to 100 GB virtio-blk qcow2 file. 3. ffsb benchmark with 8 thread I/O profiles on both raw file and qcow2 file. 4. Check that 10 virtio-blk qcow2 files, 10 GB each, are accessible and behave correctly. This ensures that more than 6 ioeventds works and that vhost-net is not impacted by adding many virtio-blk devices using ioeventfd. 5. Stephen Tweedie's verify-data on a 10 virtio-blk qcow2 file striped LVM volume. 6. Ensured that virtio-blk-pci devices are defaulting to ioeventfd=on.
Technical note added. If any revisions are required, please edit the "Technical Notes" field accordingly. All revisions will be proofread by the Engineering Content Services team. New Contents: Cause: no ioeventfd support on virtio-blk. Consequence: lower performance and scalability than what is possible if using ioeventfd. Change: ioeventfd support include on qemu-kvm virtio-blk code. Result: higher performance and scalability by decoupling guest code execution from virtio hardware emulation, reducing the amount of time stolen from the guest and allowing guest CPU and I/O to run in parallel.
Technical note updated. If any revisions are required, please edit the "Technical Notes" field accordingly. All revisions will be proofread by the Engineering Content Services team. Diffed Contents: @@ -1,7 +1 @@ -Cause: no ioeventfd support on virtio-blk. +The qemu-kvm virtio-blk code now support for ioeventfd, which results in improved performance and scalability by decoupling guest code execution from virtio hardware emulation. This decoupling allows the guest CPU and I/O to run in parallel and ensures that less time is taken from the guest.- -Consequence: lower performance and scalability than what is possible if using ioeventfd. - -Change: ioeventfd support include on qemu-kvm virtio-blk code. - -Result: higher performance and scalability by decoupling guest code execution from virtio hardware emulation, reducing the amount of time stolen from the guest and allowing guest CPU and I/O to run in parallel.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHSA-2011-0534.html
Michael, is this enabled or disabled by default? Are there cases where we think this would harm performance?
Ayal: to my knowledge it is enabled by default for virtio-blk. For other virtio devices it is not enabled by default. You can override ioeventfd by setting it on the -device virtio-net-pci|virtio-blk-pci,ioeventfd=on|off command-line.