On 6/9/20 12:23 PM, Khalid Aziz wrote:
> On Mon, 2020-06-01 at 12:48 -0700, Nitin Gupta wrote:
>> For some applications, we need to allocate almost all memory as
>> hugepages. However, on a running system, higher-order allocations can
>> fail if the memory is fragmented. Linux kernel currently does on-
>> demand
>> compaction as we request more hugepages, but this style of compaction
>> incurs very high latency. Experiments with one-time full memory
>> compaction (followed by hugepage allocations) show that kernel is
>> able
>> to restore a highly fragmented memory state to a fairly compacted
>> memory
>> state within <1 sec for a 32G system. Such data suggests that a more
>> proactive compaction can help us allocate a large fraction of memory
>> as
>> hugepages keeping allocation latencies low.
>>
>> For a more proactive compaction, the approach taken here is to define
>> a
>> new sysctl called 'vm.compaction_proactiveness' which dictates bounds
>> for external fragmentation which kcompactd tries to maintain.
>>
>> The tunable takes a value in range [0, 100], with a default of 20.
>>
>> Note that a previous version of this patch [1] was found to introduce
>> too many tunables (per-order extfrag{low, high}), but this one
>> reduces
>> them to just one sysctl. Also, the new tunable is an opaque value
>> instead of asking for specific bounds of "external fragmentation",
>> which
>> would have been difficult to estimate. The internal interpretation of
>> this opaque value allows for future fine-tuning.
>>
>> Currently, we use a simple translation from this tunable to [low,
>> high]
>> "fragmentation score" thresholds (low=100-proactiveness,
>> high=low+10%).
>> The score for a node is defined as weighted mean of per-zone external
>> fragmentation. A zone's present_pages determines its weight.
>>
>> To periodically check per-node score, we reuse per-node kcompactd
>> threads, which are woken up every 500 milliseconds to check the same.
>> If
>> a node's score exceeds its high threshold (as derived from user-
>> provided
>> proactiveness value), proactive compaction is started until its score
>> reaches its low threshold value. By default, proactiveness is set to
>> 20,
>> which implies threshold values of low=80 and high=90.
>>
>> This patch is largely based on ideas from Michal Hocko [2]. See also
>> the
>> LWN article [3].
>>
>> Performance data
>> ================
>>
>> System: x64_64, 1T RAM, 80 CPU threads.
>> Kernel: 5.6.0-rc3 + this patch
>>
>> echo madvise | sudo tee /sys/kernel/mm/transparent_hugepage/enabled
>> echo madvise | sudo tee /sys/kernel/mm/transparent_hugepage/defrag
>>
>> Before starting the driver, the system was fragmented from a
>> userspace
>> program that allocates all memory and then for each 2M aligned
>> section,
>> frees 3/4 of base pages using munmap. The workload is mainly
>> anonymous
>> userspace pages, which are easy to move around. I intentionally
>> avoided
>> unmovable pages in this test to see how much latency we incur when
>> hugepage allocations hit direct compaction.
>>
>> 1. Kernel hugepage allocation latencies
>>
>> With the system in such a fragmented state, a kernel driver then
>> allocates as many hugepages as possible and measures allocation
>> latency:
>>
>> (all latency values are in microseconds)
>>
>> - With vanilla 5.6.0-rc3
>>
>> percentile latency
>> –––––––––– –––––––
>> 5 7894
>> 10 9496
>> 25 12561
>> 30 15295
>> 40 18244
>> 50 21229
>> 60 27556
>> 75 30147
>> 80 31047
>> 90 32859
>> 95 33799
>>
>> Total 2M hugepages allocated = 383859 (749G worth of hugepages out of
>> 762G total free => 98% of free memory could be allocated as
>> hugepages)
>>
>> - With 5.6.0-rc3 + this patch, with proactiveness=20
>>
>> sysctl -w vm.compaction_proactiveness=20
>>
>> percentile latency
>> –––––––––– –––––––
>> 5 2
>> 10 2
>> 25 3
>> 30 3
>> 40 3
>> 50 4
>> 60 4
>> 75 4
>> 80 4
>> 90 5
>> 95 429
>>
>> Total 2M hugepages allocated = 384105 (750G worth of hugepages out of
>> 762G total free => 98% of free memory could be allocated as
>> hugepages)
>>
>> 2. JAVA heap allocation
>>
>> In this test, we first fragment memory using the same method as for
>> (1).
>>
>> Then, we start a Java process with a heap size set to 700G and
>> request
>> the heap to be allocated with THP hugepages. We also set THP to
>> madvise
>> to allow hugepage backing of this heap.
>>
>> /usr/bin/time
>> java -Xms700G -Xmx700G -XX:+UseTransparentHugePages
>> -XX:+AlwaysPreTouch
>>
>> The above command allocates 700G of Java heap using hugepages.
>>
>> - With vanilla 5.6.0-rc3
>>
>> 17.39user 1666.48system 27:37.89elapsed
>>
>> - With 5.6.0-rc3 + this patch, with proactiveness=20
>>
>> 8.35user 194.58system 3:19.62elapsed
>>
>> Elapsed time remains around 3:15, as proactiveness is further
>> increased.
>>
>> Note that proactive compaction happens throughout the runtime of
>> these
>> workloads. The situation of one-time compaction, sufficient to supply
>> hugepages for following allocation stream, can probably happen for
>> more
>> extreme proactiveness values, like 80 or 90.
>>
>> In the above Java workload, proactiveness is set to 20. The test
>> starts
>> with a node's score of 80 or higher, depending on the delay between
>> the
>> fragmentation step and starting the benchmark, which gives more-or-
>> less
>> time for the initial round of compaction. As t he benchmark
>> consumes
>> hugepages, node's score quickly rises above the high threshold (90)
>> and
>> proactive compaction starts again, which brings down the score to the
>> low threshold level (80). Repeat.
>>
>> bpftrace also confirms proactive compaction running 20+ times during
>> the
>> runtime of this Java benchmark. kcompactd threads consume 100% of one
>> of
>> the CPUs while it tries to bring a node's score within thresholds.
>>
>> Backoff behavior
>> ================
>>
>> Above workloads produce a memory state which is easy to compact.
>> However, if memory is filled with unmovable pages, proactive
>> compaction
>> should essentially back off. To test this aspect:
>>
>> - Created a kernel driver that allocates almost all memory as
>> hugepages
>> followed by freeing first 3/4 of each hugepage.
>> - Set proactiveness=40
>> - Note that proactive_compact_node() is deferred maximum number of
>> times
>> with HPAGE_FRAG_CHECK_INTERVAL_MSEC of wait between each check
>> (=> ~30 seconds between retries).
>>
>> [1] https://patchwork.kernel.org/patch/11098289/
>> [2]
>> https://lore.kernel.org/linux-mm/20161230131412.gi13...@dhcp22.suse.cz/
>> [3] https://lwn.net/Articles/817905/
>>
>> Signed-off-by: Nitin Gupta <nigu...@nvidia.com>
>> Reviewed-by: Vlastimil Babka <vba...@suse.cz>
>> To: Mel Gorman <mgor...@techsingularity.net>
>> To: Michal Hocko <mho...@suse.com>
>> To: Vlastimil Babka <vba...@suse.cz>
>> CC: Matthew Wilcox <wi...@infradead.org>
>> CC: Andrew Morton <a...@linux-foundation.org>
>> CC: Mike Kravetz <mike.krav...@oracle.com>
>> CC: Joonsoo Kim <iamjoonsoo....@lge.com>
>> CC: David Rientjes <rient...@google.com>
>> CC: Nitin Gupta <ngu...@nitingupta.dev>
>> CC: linux-kernel <linux-kernel@vger.kernel.org>
>> CC: linux-mm <linux...@kvack.org>
>> CC: Linux API <linux-...@vger.kernel.org>
>>
>> ---
>> Changelog v6 vs v5:
>> - Fallback to HUGETLB_PAGE_ORDER if HPAGE_PMD_ORDER is not defined,
>> and
>> some cleanups (Vlastimil)
>> - Cap min threshold to avoid excess compaction load in case user
>> sets
>> extreme values like 100 for `vm.compaction_proactiveness` sysctl
>> (Khalid)
>> - Add some more explanation about the effect of tunable on
>> compaction
>> behavior in user guide (Khalid)
>>
>> Changelog v5 vs v4:
>> - Change tunable from sysfs to sysctl (Vlastimil)
>> - Replace HUGETLB_PAGE_ORDER with HPAGE_PMD_ORDER (Vlastimil)
>> - Minor cleanups (remove redundant initializations, ...)
>>
>> Changelog v4 vs v3:
>> - Document various functions.
>> - Added admin-guide for the new tunable `proactiveness`.
>> - Rename proactive_compaction_score to fragmentation_score for
>> clarity.
>>
>> Changelog v3 vs v2:
>> - Make proactiveness a global tunable and not per-node. Also
>> upadated
>> the
>> patch description to reflect the same (Vlastimil Babka).
>> - Don't start proactive compaction if kswapd is running (Vlastimil
>> Babka).
>> - Clarified in the description that compaction runs in parallel with
>> the workload, instead of a one-time compaction followed by a
>> stream
>> of
>> hugepage allocations.
>>
>> Changelog v2 vs v1:
>> - Introduce per-node and per-zone "proactive compaction score". This
>> score is compared against watermarks which are set according to
>> user provided proactiveness value.
>> - Separate code-paths for proactive compaction from targeted
>> compaction
>> i.e. where pgdat->kcompactd_max_order is non-zero.
>> - Renamed hpage_compaction_effort -> proactiveness. In future we may
>> use more than extfrag wrt hugepage size to determine proactive
>> compaction score.
>> ---
>> Documentation/admin-guide/sysctl/vm.rst | 15 ++
>> include/linux/compaction.h | 2 +
>> kernel/sysctl.c | 9 ++
>> mm/compaction.c | 183
>> +++++++++++++++++++++++-
>> mm/internal.h | 1 +
>> mm/vmstat.c | 18 +++
>> 6 files changed, 223 insertions(+), 5 deletions(-)
>
>
> Looks good to me.
>
> Reviewed-by: Khalid Aziz <khalid.a...@oracle.com>
>
>
Thanks Khalid.
Andrew, this patch now has two 'Reviewed-by' (Vlastimil and Khalid).
Can this be considered for 5.8 merge now?
Thanks,
Nitin
