Re: [PATCH v28 0/4] Virtio-balloon: support free page reporting
On 02/09/2018 11:14 AM, Michael S. Tsirkin wrote: On Fri, Feb 09, 2018 at 11:11:39AM +0800, Wei Wang wrote: On 02/09/2018 03:55 AM, Michael S. Tsirkin wrote: On Thu, Feb 08, 2018 at 05:50:16PM +0800, Wei Wang wrote: Details: Set up a Ping-Pong local live migration, where the guest ceaselessy migrates between the source and destination. Linux compilation, i.e. make bzImage -j4, is performed during the Ping-Pong migration. The legacy case takes 5min14s to finish the compilation. With this optimization patched, it takes 5min12s. How is migration time affected in this case? When the linux compilation workload runs, the migration time (both the legacy and this optimization case) varies as the compilation goes on. It seems not easy to give a static speedup number, some times the migration time is reduced to 33%, sometimes to 50%, it varies, and depends on how much free memory the system has at that moment. For example, at the later stage of the compilation, I can observe 5GB memory being used as page cache. But overall, I can observe obvious improvement of the migration time. Best, Wei You can run multiple tests and give a best, worst and median numbers. Sorry for my late response (I was on leaves for some other responsibilities). Here are some more numbers of the live migration time comparison while linux compilation is in progress (probably the average time is more relevant) average: 52.4% reduction (optimization v.s. legacy = 1242ms v.s. 2611ms) best: 69.1% worst: 21.9% Best, Wei
Re: [PATCH v28 0/4] Virtio-balloon: support free page reporting
On Fri, Feb 09, 2018 at 11:11:39AM +0800, Wei Wang wrote: > On 02/09/2018 03:55 AM, Michael S. Tsirkin wrote: > > On Thu, Feb 08, 2018 at 05:50:16PM +0800, Wei Wang wrote: > > > > > Details: > > > Set up a Ping-Pong local live migration, where the guest ceaselessy > > > migrates between the source and destination. Linux compilation, > > > i.e. make bzImage -j4, is performed during the Ping-Pong migration. The > > > legacy case takes 5min14s to finish the compilation. With this > > > optimization patched, it takes 5min12s. > > How is migration time affected in this case? > > > When the linux compilation workload runs, the migration time (both the > legacy and this optimization case) varies as the compilation goes on. It > seems not easy to give a static speedup number, some times the migration > time is reduced to 33%, sometimes to 50%, it varies, and depends on how much > free memory the system has at that moment. For example, at the later stage > of the compilation, I can observe 5GB memory being used as page cache. But > overall, I can observe obvious improvement of the migration time. > > > Best, > Wei You can run multiple tests and give a best, worst and median numbers. -- MST
Re: [PATCH v28 0/4] Virtio-balloon: support free page reporting
On 02/09/2018 03:55 AM, Michael S. Tsirkin wrote: On Thu, Feb 08, 2018 at 05:50:16PM +0800, Wei Wang wrote: Details: Set up a Ping-Pong local live migration, where the guest ceaselessy migrates between the source and destination. Linux compilation, i.e. make bzImage -j4, is performed during the Ping-Pong migration. The legacy case takes 5min14s to finish the compilation. With this optimization patched, it takes 5min12s. How is migration time affected in this case? When the linux compilation workload runs, the migration time (both the legacy and this optimization case) varies as the compilation goes on. It seems not easy to give a static speedup number, some times the migration time is reduced to 33%, sometimes to 50%, it varies, and depends on how much free memory the system has at that moment. For example, at the later stage of the compilation, I can observe 5GB memory being used as page cache. But overall, I can observe obvious improvement of the migration time. Best, Wei
Re: [PATCH v28 0/4] Virtio-balloon: support free page reporting
On Thu, Feb 08, 2018 at 05:50:16PM +0800, Wei Wang wrote: > This patch series is separated from the previous "Virtio-balloon > Enhancement" series. The new feature, VIRTIO_BALLOON_F_FREE_PAGE_HINT, > implemented by this series enables the virtio-balloon driver to report > hints of guest free pages to the host. It can be used to accelerate live > migration of VMs. Here is an introduction of this usage: > > Live migration needs to transfer the VM's memory from the source machine > to the destination round by round. For the 1st round, all the VM's memory > is transferred. From the 2nd round, only the pieces of memory that were > written by the guest (after the 1st round) are transferred. One method > that is popularly used by the hypervisor to track which part of memory is > written is to write-protect all the guest memory. > > This feature enables the optimization of the 1st round memory transfer - > the hypervisor can skip the transfer of guest free pages in the 1st round. > It is not concerned that the memory pages are used after they are given > to the hypervisor as a hint of the free pages, because they will be > tracked by the hypervisor and transferred in the next round if they are > used and written. At this point I think it's a good idea to focus on the qemu side, that code is still not really ready yet. > * Tests > - Migration time improvement > Result: > Live migration time is reduced to 14% with this optimization. > Details: > Local live migration of 8GB idle guest, the legacy live migration takes > ~1817ms. With this optimization, it takes ~254ms, which reduces the time > to 14%. > - Workload tests > Results: > Running this feature has no impact on the linux compilation workload > running inside the guest. I think you should try something memory intensive. Try asking qemu migration guys for hints on a good test to run. > Details: > Set up a Ping-Pong local live migration, where the guest ceaselessy > migrates between the source and destination. Linux compilation, > i.e. make bzImage -j4, is performed during the Ping-Pong migration. The > legacy case takes 5min14s to finish the compilation. With this > optimization patched, it takes 5min12s. How is migration time affected in this case? > ChangeLog: > v27->v28: > - mm/page_poison: Move PAGE_POISON to page_poison.c and add a function > to expose page poison val to kernel modules. > v26->v27: > - add a new patch to expose page_poisoning_enabled to kernel modules > - virtio-balloon: set poison_val to 0x, instead of 0xaa > v25->v26: virtio-balloon changes only > - remove kicking free page vq since the host now polls the vq after > initiating the reporting > - report_free_page_func: detach all the used buffers after sending > the stop cmd id. This avoids leaving the detaching burden (i.e. > overhead) to the next cmd id. Detaching here isn't considered > overhead since the stop cmd id has been sent, and host has already > moved formard. > v24->v25: > - mm: change walk_free_mem_block to return 0 (instead of true) on > completing the report, and return a non-zero value from the > callabck, which stops the reporting. > - virtio-balloon: > - use enum instead of define for VIRTIO_BALLOON_VQ_INFLATE etc. > - avoid __virtio_clear_bit when bailing out; > - a new method to avoid reporting the some cmd id to host twice > - destroy_workqueue can cancel free page work when the feature is > negotiated; > - fail probe when the free page vq size is less than 2. > v23->v24: > - change feature name VIRTIO_BALLOON_F_FREE_PAGE_VQ to > VIRTIO_BALLOON_F_FREE_PAGE_HINT > - kick when vq->num_free < half full, instead of "= half full" > - replace BUG_ON with bailing out > - check vb->balloon_wq in probe(), if null, bail out > - add a new feature bit for page poisoning > - solve the corner case that one cmd id being sent to host twice > v22->v23: > - change to kick the device when the vq is half-way full; > - open-code batch_free_page_sg into add_one_sg; > - change cmd_id from "uint32_t" to "__virtio32"; > - reserver one entry in the vq for the driver to send cmd_id, instead > of busywaiting for an available entry; > - add "stop_update" check before queue_work for prudence purpose for > now, will have a separate patch to discuss this flag check later; > - init_vqs: change to put some variables on stack to have simpler > implementation; > - add destroy_workqueue(vb->balloon_wq); > > v21->v22: > - add_one_sg: some code and comment re-arrangement > - send_cmd_id: handle a cornercase > > For previous ChangeLog, please reference > https://lwn.net/Articles/743660/ > > Wei Wang (4): > mm: support reporting free page blocks > virtio-balloon: VIRTIO_BALLOON_F_FREE_PAGE_HINT > mm/page_poison: add a function to expose page poison val to kernel
[PATCH v28 0/4] Virtio-balloon: support free page reporting
This patch series is separated from the previous "Virtio-balloon Enhancement" series. The new feature, VIRTIO_BALLOON_F_FREE_PAGE_HINT, implemented by this series enables the virtio-balloon driver to report hints of guest free pages to the host. It can be used to accelerate live migration of VMs. Here is an introduction of this usage: Live migration needs to transfer the VM's memory from the source machine to the destination round by round. For the 1st round, all the VM's memory is transferred. From the 2nd round, only the pieces of memory that were written by the guest (after the 1st round) are transferred. One method that is popularly used by the hypervisor to track which part of memory is written is to write-protect all the guest memory. This feature enables the optimization of the 1st round memory transfer - the hypervisor can skip the transfer of guest free pages in the 1st round. It is not concerned that the memory pages are used after they are given to the hypervisor as a hint of the free pages, because they will be tracked by the hypervisor and transferred in the next round if they are used and written. * Tests - Migration time improvement Result: Live migration time is reduced to 14% with this optimization. Details: Local live migration of 8GB idle guest, the legacy live migration takes ~1817ms. With this optimization, it takes ~254ms, which reduces the time to 14%. - Workload tests Results: Running this feature has no impact on the linux compilation workload running inside the guest. Details: Set up a Ping-Pong local live migration, where the guest ceaselessy migrates between the source and destination. Linux compilation, i.e. make bzImage -j4, is performed during the Ping-Pong migration. The legacy case takes 5min14s to finish the compilation. With this optimization patched, it takes 5min12s. ChangeLog: v27->v28: - mm/page_poison: Move PAGE_POISON to page_poison.c and add a function to expose page poison val to kernel modules. v26->v27: - add a new patch to expose page_poisoning_enabled to kernel modules - virtio-balloon: set poison_val to 0x, instead of 0xaa v25->v26: virtio-balloon changes only - remove kicking free page vq since the host now polls the vq after initiating the reporting - report_free_page_func: detach all the used buffers after sending the stop cmd id. This avoids leaving the detaching burden (i.e. overhead) to the next cmd id. Detaching here isn't considered overhead since the stop cmd id has been sent, and host has already moved formard. v24->v25: - mm: change walk_free_mem_block to return 0 (instead of true) on completing the report, and return a non-zero value from the callabck, which stops the reporting. - virtio-balloon: - use enum instead of define for VIRTIO_BALLOON_VQ_INFLATE etc. - avoid __virtio_clear_bit when bailing out; - a new method to avoid reporting the some cmd id to host twice - destroy_workqueue can cancel free page work when the feature is negotiated; - fail probe when the free page vq size is less than 2. v23->v24: - change feature name VIRTIO_BALLOON_F_FREE_PAGE_VQ to VIRTIO_BALLOON_F_FREE_PAGE_HINT - kick when vq->num_free < half full, instead of "= half full" - replace BUG_ON with bailing out - check vb->balloon_wq in probe(), if null, bail out - add a new feature bit for page poisoning - solve the corner case that one cmd id being sent to host twice v22->v23: - change to kick the device when the vq is half-way full; - open-code batch_free_page_sg into add_one_sg; - change cmd_id from "uint32_t" to "__virtio32"; - reserver one entry in the vq for the driver to send cmd_id, instead of busywaiting for an available entry; - add "stop_update" check before queue_work for prudence purpose for now, will have a separate patch to discuss this flag check later; - init_vqs: change to put some variables on stack to have simpler implementation; - add destroy_workqueue(vb->balloon_wq); v21->v22: - add_one_sg: some code and comment re-arrangement - send_cmd_id: handle a cornercase For previous ChangeLog, please reference https://lwn.net/Articles/743660/ Wei Wang (4): mm: support reporting free page blocks virtio-balloon: VIRTIO_BALLOON_F_FREE_PAGE_HINT mm/page_poison: add a function to expose page poison val to kernel modules virtio-balloon: VIRTIO_BALLOON_F_PAGE_POISON drivers/virtio/virtio_balloon.c | 258 +++- include/linux/mm.h | 8 ++ include/linux/poison.h | 7 - include/uapi/linux/virtio_balloon.h | 7 + mm/page_alloc.c | 96 ++ mm/page_poison.c| 24 6 files changed, 357 insertions(+), 43 deletions(-) -- 2.7.4
