Hi Pengcheng,

I haven't gone that far yet. I am still looking for a way to saturate this
setup and realistic numbers for this new workload.

With Olio, I had about 10 client machines, 1 front-end, 5 backends (using
haproxy). That setup allowed me to put up to 6K concurrent users on a
front-end without violating SLA. Note that I had very powerful servers with
8 network cards on each server using IP bonding and also 8 disks. This new
wokload uses memcache which may reduce the frontend and backend load.

I am running in separated machines because I want to do micro-architectural
characterization of these workloads. Thay is why I split the services.

On Thursday, 3 March 2016, Pengcheng Li <[email protected]> wrote:

> Leonardo,
>
> So based on your operation steps, so there is no any way to saturate this
> network using one node of client? Right?
>
> Best,
> Pengcheng
>
> On Mar 2, 2016, at 11:11 PM, Leonardo Piga <[email protected]
> <javascript:_e(%7B%7D,'cvml','[email protected]');>> wrote:
>
> Thanks Alex. I followed that steps and got it to work with multiple
> machines.
>
> Now, what is your recommendation to increase the load of the web server?
> How many users and fa:scale value are you guys using to saturate the web
> server?
>
> The first step of run.sh generates about 500 users. This seems pretty low
> for me. I remember to have more than 4K concurrent users in Olio to
> saturate the web server.
>
> Regards,
> --
> Leonardo
>
> On Monday, 29 February 2016, Alexandros Daglis <[email protected]
> <javascript:_e(%7B%7D,'cvml','[email protected]');>> wrote:
>
>> Hi Leonardo,
>>
>> We also provide an example of a multinode Memcached deployment, using
>> Docker Swarm, here
>> <https://urldefense.proofpoint.com/v2/url?u=http-3A__cloudsuite.ch_swarm_&d=BQMFaQ&c=kbmfwr1Yojg42sGEpaQh5ofMHBeTl9EI2eaqQZhHbOU&r=URNx93oZUKClkw67__wKyT0riNmKMq5966ML77bMvbs&m=CsNu9cx1XohOMzv-EwDnKOUovQZot9gW0WwE_N2wYqw&s=rx7BqZw-0S3K3rp2QpMoLO4U09LSlfewrLZUI7K3yyg&e=>.
>> It should be easy to follow the same process for Web Serving.
>>
>> Regards,
>> Alex
>>
>> On Feb 29, 2016, at 7:32 PM, Nooshin Mirzadeh <[email protected]>
>> wrote:
>>
>> Dear Leonardo,
>>
>> Thank you very much for your email, and the fixes. I have applied the
>> changes, and we will push them to the master branch soon.
>> The warnings are fine. There is no problem.
>>
>> You can run them on different machines by using the host network (instead
>> of using my_network that is created). Also, you can take a look at Docker
>> Swarm
>> <https://urldefense.proofpoint.com/v2/url?u=https-3A__docs.docker.com_swarm_overview_&d=BQMFaQ&c=kbmfwr1Yojg42sGEpaQh5ofMHBeTl9EI2eaqQZhHbOU&r=URNx93oZUKClkw67__wKyT0riNmKMq5966ML77bMvbs&m=CsNu9cx1XohOMzv-EwDnKOUovQZot9gW0WwE_N2wYqw&s=PChzFH5bVgKqfIzaCC-vno8GwEXGsEiHINRUdxV0TV0&e=>
>> which is for multi-node networking. It enables you to create cluster of
>> containers over multiple nodes and creates an abstraction for you. In the
>> future version of CloudSuite, we may consider it, as well.
>>
>> Regards,
>> Nooshin
>>
>>
>>
>> On Sat, Feb 27, 2016 at 9:33 AM, Leonardo Piga <[email protected]>
>> wrote:
>>
>>> Hi all,
>>>
>>> I was following the instructions to get the new Web Serving benchmark
>>> from here:
>>> http://cloudsuite.ch/webserving/
>>> <https://urldefense.proofpoint.com/v2/url?u=http-3A__cloudsuite.ch_webserving_&d=BQMFaQ&c=kbmfwr1Yojg42sGEpaQh5ofMHBeTl9EI2eaqQZhHbOU&r=URNx93oZUKClkw67__wKyT0riNmKMq5966ML77bMvbs&m=CsNu9cx1XohOMzv-EwDnKOUovQZot9gW0WwE_N2wYqw&s=mS_k35IKARqQMnrV2GmdaRRd2RcfuNB6q_-QzVjke04&e=>
>>>
>>> As a CloudSuite user since version 1, I appreciate the effort of using
>>> Dockers to make the users' life easier. However, when things do not
>>> work properly, it is very painful to debug without properly
>>> instruction. I used some of my past experience to track the following
>>> issues down.
>>>
>>> I don't know if you guys never tested the Web Serving Dockers or if
>>> you changed something after testing and broke things (most probably).
>>> After some hours reverse engineering the Dockers and installations, I
>>> finally manage to make it work.
>>>
>>> Here is the list of things that need to be fixed:
>>>
>>> 1) The web server docker should be named php_server, otherwise many
>>> things will break.I figured out that the db_server docker is still
>>> using php_server to refer to the front-end. This causes many problems
>>> with redirectiong. In the instructions, change
>>>
>>> $ docker run -d -t --net=my_net --privileged=true --name=web_server
>>> cloudsuite/web-serving:web_server /etc/bootstrap.sh
>>>
>>> to
>>>
>>> $ docker run -d -t --net=my_net --privileged=true --name=php_server
>>> cloudsuite/web-serving:web_server /etc/bootstrap.sh
>>>
>>>
>>> 2) Fix the run.sh script to look for php_server instead of web_server.
>>> Change
>>>
>>> docker exec faban_client sh -c "cd /web20_benchmark/build && java -jar
>>> Usergen.jar http://$
>>> <https://urldefense.proofpoint.com/v2/url?u=http-3A__-24&d=BQMFaQ&c=kbmfwr1Yojg42sGEpaQh5ofMHBeTl9EI2eaqQZhHbOU&r=URNx93oZUKClkw67__wKyT0riNmKMq5966ML77bMvbs&m=CsNu9cx1XohOMzv-EwDnKOUovQZot9gW0WwE_N2wYqw&s=wkFavt7EWndpgeX8R-Cr1J5XtyWETbnc84HZEDThB6c&e=>(docker
>>> inspect --format '{{
>>> .NetworkSettings.Networks.my_net.IPAddress }}' web_server):8080"
>>>
>>> to
>>>
>>> docker exec faban_client sh -c "cd /web20_benchmark/build && java -jar
>>> Usergen.jar http://$
>>> <https://urldefense.proofpoint.com/v2/url?u=http-3A__-24&d=BQMFaQ&c=kbmfwr1Yojg42sGEpaQh5ofMHBeTl9EI2eaqQZhHbOU&r=URNx93oZUKClkw67__wKyT0riNmKMq5966ML77bMvbs&m=CsNu9cx1XohOMzv-EwDnKOUovQZot9gW0WwE_N2wYqw&s=wkFavt7EWndpgeX8R-Cr1J5XtyWETbnc84HZEDThB6c&e=>(docker
>>> inspect --format '{{
>>> .NetworkSettings.Networks.my_net.IPAddress }}' php_server):8080"
>>>
>>> run.sh is pretty inconsistent, first half uses  web_server and other
>>> half uses php_server.
>>>
>>> 3) There is a nasty bug in the file
>>> /usr/share/nginx/html/elgg/engine/setting.php in the web server docker
>>> that prevents memcache to work properly. In this file change
>>>
>>> $CONFIG->memcache_servers = array (
>>> //      array('server1', 11211),
>>> //      array('10.22.17.71', 11211)
>>>     array('memcache_server, 11211')
>>> );
>>>
>>> to
>>>
>>> $CONFIG->memcache_servers = array (
>>> //      array('server1', 11211),
>>> //      array('10.22.17.71', 11211)
>>>     array('memcache_server', 11211)
>>> );
>>>
>>> With these changes I was able to run faban_client. However, I still
>>> get many warnings. Please find attached the output of the run.sh. Do
>>> you think I should worry about these warnings?
>>>
>>> Also, do you have any suggestion to run the different servers and the
>>> clients on different machines? Perhaps using docker-machines? Let me
>>> know if you are planning to provide instructions in this lines.
>>>
>>> Regards,
>>> --
>>> Leonardo
>>>
>>
>>
>>
>
> --
> Leonardo
>
>
>

-- 
Leonardo

Reply via email to