Re: [RFC] implement QUEUED spinlocks on powerpc

[panxinhui <xinhui@linux.vnet.ibm.com>]

[-- Attachment #1: Type: text/plain, Size: 2418 bytes --]



在 2017/2/13 下午5:08, panxinhui 写道:
> 
> 
> 在 2017/2/7 下午2:46, Eric Dumazet 写道:
>> On Mon, Feb 6, 2017 at 10:21 PM, panxinhui <xinhui@linux.vnet.ibm.com> wrote:
>>
>>> hi all
>>>         I do some netperf tests and get some benchmark results.
>>> I also attach my test script and netperf-result(Excel)
>>>
> HI, all
> I use loopback interface to run netperf tests,
> #tc qd add dev lo root pfifo limit 10000
> #ip link
> 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc pfifo state UNKNOWN mode DEFAULT group default qlen 1000
>     link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
> 
> and put the result in netperf.xlsx(excel)
> 
> It is a 32 vcpus P8 machine, with 32Gib memory.
> 
> This time spinlock is the best one, qspinlock > pvqspinlock. So sad.
> 
This time, I have appiled some optimising patches on pvqspinlock.
When there is a high contention, the performance has a good improvement ans is very similar to spinlock.

Result is attached in netperf.xlsx

thanks
xinhui

> thanks
> xinhui
>>> There are two machine. one runs netserver and the other runs netperf
>>> benchmark. 1000Mbps network is connected with them.
>>>
>>> #ip link infomation
>>> 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state
>>> UNKNOWN mode DEFAULT group default qlen 1000
>>>      link/ether ba:68:9c:14:32:02 brd ff:ff:ff:ff:ff:ff
>>>
>>> According to the results, there is not much performance gap with each other.
>>> And as we are only testing the throughput, the pvqspinlock shows the
>>> overhead of its pv stuff. but qspinlock shows a little improvement than
>>> spinlock. My simple summary in this testcase is
>>> qspinlock > spinlock > pvqspinlock.
>>>
>>> when run 200 concurrent netperf, I paste the total throughput here.
>>>
>>>         concurrent runners| total throughput | variance
>>> -------------------------------------------
>>> spinlock        | 199 | 66882.8 | 89.93
>>> -------------------------------------------
>>> qspinlock       | 199 | 66350.4 | 72.0239
>>> -------------------------------------------
>>> pvqspinlock     | 199 | 64740.5 | 85.7837
>>>
>>> You could see more data in nerperf.xlsx
>>>
>>> thanks
>>> xinhui
>>
>>
>> Hi xinhui
>>
>> 1Gbit NIC is too slow for this use case. I would try a 10Gbit NIC at least...
>>
>> Alternatively, you could use loopback interface.  (netperf -H 127.0.0.1)
>>
>> tc qd add dev lo root pfifo limit 10000
>>

[-- Attachment #2: netperf.xlsx --]
[-- Type: application/vnd.openxmlformats-officedocument.spreadsheetml.sheet, Size: 72195 bytes --]