[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <c8e771c8-4b01-f2b4-5b54-e9931f556270@intel.com>
Date: Thu, 8 Sep 2022 16:25:47 +0800
From: "Sun, Jiebin" <jiebin.sun@...el.com>
To: Andrew Morton <akpm@...ux-foundation.org>,
Tim Chen <tim.c.chen@...ux.intel.com>
Cc: vasily.averin@...ux.dev, shakeelb@...gle.com, dennis@...nel.org,
tj@...nel.org, cl@...ux.com, ebiederm@...ssion.com,
legion@...nel.org, manfred@...orfullife.com,
alexander.mikhalitsyn@...tuozzo.com, linux-mm@...ck.org,
linux-kernel@...r.kernel.org, tim.c.chen@...el.com,
feng.tang@...el.com, ying.huang@...el.com, tianyou.li@...el.com,
wangyang.guo@...el.com, jiebin.sun@...el.com
Subject: Re: [PATCH v4] ipc/msg: mitigate the lock contention with percpu
counter
On 9/8/2022 5:34 AM, Andrew Morton wrote:
> On Wed, 07 Sep 2022 09:01:53 -0700 Tim Chen <tim.c.chen@...ux.intel.com> wrote:
>
>> On Thu, 2022-09-08 at 01:25 +0800, Jiebin Sun wrote:
>>> The msg_bytes and msg_hdrs atomic counters are frequently
>>> updated when IPC msg queue is in heavy use, causing heavy
>>> cache bounce and overhead. Change them to percpu_counter
>>> greatly improve the performance. Since there is one percpu
>>> struct per namespace, additional memory cost is minimal.
>>> Reading of the count done in msgctl call, which is infrequent.
>>> So the need to sum up the counts in each CPU is infrequent.
>>>
>>>
>>> Apply the patch and test the pts/stress-ng-1.4.0
>>> -- system v message passing (160 threads).
>>>
>>> Score gain: 3.17x
>>>
>>>
>> ...
>>>
>>> +/* large batch size could reduce the times to sum up percpu counter */
>>> +#define MSG_PERCPU_COUNTER_BATCH 1024
>>> +
>> Jiebin,
>>
>> 1024 is a small size (1/4 page).
>> The local per cpu counter could overflow to the gloabal count quickly
>> if it is limited to this size, since our count tracks msg size.
>>
>> I'll suggest something larger, say 8*1024*1024, about
>> 8MB to accommodate about 2 large page worth of data. Maybe that
>> will further improve throughput on stress-ng by reducing contention
>> on adding to the global count.
>>
> I think this concept of a percpu_counter_add() which is massively
> biased to the write side and with very rare reading is a legitimate
> use-case. Perhaps it should become an addition to the formal interface.
> Something like
>
> /*
> * comment goes here
> */
> static inline void percpu_counter_add_local(struct percpu_counter *fbc,
> s64 amount)
> {
> percpu_counter_add_batch(fbc, amount, INT_MAX);
> }
>
> and percpu_counter_sub_local(), I guess.
>
> The only instance I can see is
> block/blk-cgroup-rwstat.h:blkg_rwstat_add() which is using INT_MAX/2
> because it always uses percpu_counter_sum_positive() on the read side.
>
> But that makes two!
Yes. Using INT_MAX or INT_MAX/2 could have a big improvement on the
performance if heavy writing but rare reading. In our case, if the local
percpu counter is near to INT_MAX and there comes a big msgsz, the
overflow issue could happen. So I think INT_MAX/2, which is used in
blkg_rwstat_add(), might be a better choice. /$
percpu_counter_add_batch(&ns->percpu_msg_bytes, msgsz, batch); /I will
send the performance data and draft patch out for discussing.//Jiebin//
Powered by blists - more mailing lists