Internships/ProjectIdeas/NVMePerformance: Difference between revisions

From QEMU
(Created page with " === QEMU NVMe Performance Optimization === '''Summary:''' QEMU's NVMe implementation uses traditional trap-and-emulation method to emulate I/Os, thus the performance suffer...")
 
 
(3 intermediate revisions by the same user not shown)
Line 1: Line 1:
=== NVMe Emulation Performance Optimization ===
'''Summary:'''
QEMU's NVMe emulation uses the traditional trap-and-emulate method to
emulate I/Os, thus the performance suffers due to frequent VM-exits.
Version 1.3 of the NVMe specification defines a new feature to update
doorbell registers using a Shadow Doorbell Buffer. This can be utilized
to enhance performance of emulated controllers by reducing the number of
Submission Queue Tail Doorbell writes.


=== QEMU NVMe Performance Optimization ===
Further more, it is possible to run emulation in a dedicated thread
'''Summary:''' 
called an IOThread. Emulating NVMe in a separate thread allows the vcpu
QEMU's NVMe implementation uses traditional trap-and-emulation method to emulate I/Os, thus the performance suffers due to frequent VM-exits. NVMe Specification 1.3 defines a new feature to update doorbell registers using a Shadow Doorbell Buffer. This can be utilized to enhance performance of emulated controllers like QEMU NVMe. The goal of this summer of code is to add such support in QEMU and apply polling techniques to achieve comparable performance as virtio-blk dataplane. Specifically, this project includes the following parts: (1) add shadow doorbell buffer and ioeventfd support into QEMU NVMe emulation, which will reduce the number of VM-exits and make them less expensive (reducing VCPU latency); (2) add iothread support to QEMU NVMe emulation to reduce or eliminate VM-exits; (3) add a RAM disk back-end for debugging; (4) implement an interrupt coalescing scheme for efficient host-to-guest communication (at least one of (3) and (4)).
thread to continue execution and results in better performance.
 
Finally, it is possible for the emulation code to watch for changes to
the queue memory instead of waiting for doorbell writes. This technique
is called polling and reduces notification latency at the expense of an
another thread consuming CPU to detect queue activity.
 
The goal of this project is to add implement these optimizations so
QEMU's NVMe emulation performance becomes comparable to virtio-blk
performance.
 
Tasks include:
* Add Shadow Doorbell Buffer support to reduce doorbell writes
* Add Submission Queue Tail Doorbell register ioeventfd support when the Shadow Doorbell Buffer is enabled (see existing patch linked below)
* Add Submission Queue polling
* Add IOThread support so emulation can run in a dedicated thread


'''Links:'''
'''Links:'''
Line 9: Line 32:
* https://github.com/ucare-uchicago/femu
* https://github.com/ucare-uchicago/femu
* https://vmsplice.net/~stefan/stefanha-kvm-forum-2017.pdf
* https://vmsplice.net/~stefan/stefanha-kvm-forum-2017.pdf
* https://patchwork.kernel.org/patch/10259575/
* https://lore.kernel.org/qemu-devel/1447825624-17011-1-git-send-email-mlin@kernel.org/T/#u
'''Details:'''
'''Details:'''
* Skill level: intermediate-advanced
* Project size: 350 hours
* Language: C
* Difficulty: intermediate to advanced
* Mentor: Paolo Bonzini <pbonzini@redhat.com>, bonzini on IRC
* Required skills: C programming
* Suggested by: Huaicheng Li <huaicheng@cs.uchicago.edu>, Paolo Bonzini
* Desirable skills: knowledge of the NVMe PCI specification, knowledge of device driver or emulator development
* Mentor: Klaus Jensen <its@irrelevant.dk> (kjensen on IRC), Keith Busch <kbusch@kernel.org>
* Suggested by: Huaicheng Li <huaicheng@cs.uchicago.edu>, Paolo Bonzini <pbonzini@redhat.com> ("bonzini" on IRC)

Latest revision as of 12:38, 25 February 2022

NVMe Emulation Performance Optimization

Summary: QEMU's NVMe emulation uses the traditional trap-and-emulate method to emulate I/Os, thus the performance suffers due to frequent VM-exits. Version 1.3 of the NVMe specification defines a new feature to update doorbell registers using a Shadow Doorbell Buffer. This can be utilized to enhance performance of emulated controllers by reducing the number of Submission Queue Tail Doorbell writes.

Further more, it is possible to run emulation in a dedicated thread called an IOThread. Emulating NVMe in a separate thread allows the vcpu thread to continue execution and results in better performance.

Finally, it is possible for the emulation code to watch for changes to the queue memory instead of waiting for doorbell writes. This technique is called polling and reduces notification latency at the expense of an another thread consuming CPU to detect queue activity.

The goal of this project is to add implement these optimizations so QEMU's NVMe emulation performance becomes comparable to virtio-blk performance.

Tasks include:

  • Add Shadow Doorbell Buffer support to reduce doorbell writes
  • Add Submission Queue Tail Doorbell register ioeventfd support when the Shadow Doorbell Buffer is enabled (see existing patch linked below)
  • Add Submission Queue polling
  • Add IOThread support so emulation can run in a dedicated thread

Links:

Details:

  • Project size: 350 hours
  • Difficulty: intermediate to advanced
  • Required skills: C programming
  • Desirable skills: knowledge of the NVMe PCI specification, knowledge of device driver or emulator development
  • Mentor: Klaus Jensen <its@irrelevant.dk> (kjensen on IRC), Keith Busch <kbusch@kernel.org>
  • Suggested by: Huaicheng Li <huaicheng@cs.uchicago.edu>, Paolo Bonzini <pbonzini@redhat.com> ("bonzini" on IRC)
Retrieved from "https://wiki.qemu.org/index.php?title=Internships/ProjectIdeas/NVMePerformance&oldid=10797"