number of users supported by the Kurento Room Demo

[ya...@streaminggalicia.com]

Hi,

I was wondering if there is any test in the Kurento Room Demo project that could be used to measure the number of users supported by the system.

My idea is to simulate a scenario with many users connected via WebRTC to the application.

Thank you

[Ivan Gracia]

Perhaps you could use this one. Have a look in that test project, and you are bound to find something, or just get the hang on how we create tests, in order to replicate those.

Ivan Gracia

--
You received this message because you are subscribed to the Google Groups "kurento" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kurento+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[ya...@streaminggalicia.com]

I am using an older version of the Kurento Room so I am using the attached test.

When I run the test the output in the console is the attached "output.txt" file.

Could anyone help me?

Thanks.

[Ivan Gracia]

It looks like you forgot to create the room before joining in.

Ivan Gracia

--

[ya...@streaminggalicia.com]

Thanks, but now it shows the error in the attached file. Do you know how to solve it?

[Ivan Gracia]

Looks like it can’t connect to the room server

Caused by: org.kurento.commons.exception.KurentoException:  Timeout of 15000ms when waiting to connect to Websocket server ws://127.0.0.1:8080/room

I think you are trying to run the tests, right? You can skip them with -DskipTests.

​

Ivan Gracia

On Mon, Apr 18, 2016 at 7:38 PM, <ya...@streaminggalicia.com> wrote:

Thanks, but now it shows the error in the attached file. Do you know how to solve it?

--

[ya...@streaminggalicia.com]

Yes, I'm trying to run the test to estimate the maximum number of users the Kurento-room-demo supports. I'm interested in knowing this number. 

Do you know any other method to achieve it?

[Ivan Gracia]

Perhaps Radu can give you a hand here.

@Radu, do you know what might be happening? Yago is trying to make a load test, based on the room tests.

Ivan Gracia

On Tue, Apr 19, 2016 at 8:09 PM, <ya...@streaminggalicia.com> wrote:

Yes, I'm trying to run the test to estimate the maximum number of users the Kurento-room-demo supports. I'm interested in knowing this number. 

Do you know any other method to achieve it?

--

[Radu Tom Vlad]

Hi, from the last attachment it seems that the Room Demo app isn't starting correctly by the test setup. Please specify the commit's hash for the code that you're trying to benchmark so that I can look into it.. Also, can you rerun the test with the log level of 'org.kurento.room' set to DEBUG?

The maximum number of concurrent users for a WebRTC room-type application depends on many conditions such as hardware configuration and system load (both on the server as on the clients), network bandwidth and latency, etc.. Even though the test you're trying to run could give you an approximate on the number of supported users in a room, I'd still test it under real-world conditions (e.g. invite a bunch of people from different geographic locations and with different network access to join your room).

Please note that the demo application is just that, a demo, so high concurrency wasn't a major factor during its development phase. 

As a last thought, we'd appreciate a lot if you reported any additional bugs encountered during your tests.

Best regards,

Radu

[ya...@streaminggalicia.com]

Thank you both for the quickly response and for the information.

I just rerun the test with the log level of 'org.kurento.room' set to DEBUG and I think the output (in the attached file) is the same than in the other case.

Regarding this: "Please specify the commit's hash for the code that you're trying to benchmark so that I can look into it.". Where is the information you are asking me for?

[Radu Tom Vlad]

Hmm it's kinda strange that the log didn't change.. Have you set the log level in the log4j config file from the test project's src/test/resources?

The commit hash for your current revision (HEAD) can be obtained like this:

git rev-parse HEAD

[ya...@streaminggalicia.com]

Yes, I have set the log level in the log4j config file from the test project's src/test/resources. I attach an image of the log configuration.

The command 'git rev-parse HEAD' run under the kurento-room folder shows this output:

e74ebe675c3e41188bc47c873536efbdca1dbbc7

 

[Radu Tom Vlad]

I'm sorry but I can't seem to be able to find the commit by the hash that you've indicated.. Could you please run git log and copy the first lines of the resulting output?

[ya...@streaminggalicia.com]

Well, I've downloaded again the Kurento-Room-6.2.0 and it works.

I've run this test with only two users to check wether the room works or not. The test could be run with different number of users so how can I figure out the maximum number?

I attach the output for two users.

[Radu Tom Vlad]

I'm glad that solved it for you. You could try to configure more fake users in order to get to the max supported user sessions.

You have to take into account that the KMS used by these users (kurento.test.fake.kmsUri) can quickly become the bottleneck of the whole system once you start adding more and more users. It's advisable for it to point to a separate KMS instance (from the one used by the room demo) and you could also modify the test in order to distribute the fake users onto several KMS instances.

[ya...@streaminggalicia.com]

Thanks for your help.

I have another question related to that. Do you know which parameters of the wertc-internals may indicate a degrade in the videoconferencing quality? There are a lot of parameters. I guess the RTT and the lost packets are some of them but I want to know the others if it is possible. I'm also interesting in the range this values have to take in order to perform a good quality videoconferencing.

I attach two images for both the audio and video stats of the sender.

Thank you all

[Radu Tom Vlad]

Hi again, I'm not really sure which of these params are most suitable for your analysis.. I think packetsLost or bitsSentPerSecond would let you know if you have drops in the network traffic that then might translate into a quality degradation of the video or audio stream. If you want more specific details I'd consult with the WebRTC groups like discuss-webrtc, they could provide additional info.

Regards,

Radu

[ya...@streaminggalicia.com]

Thank you. I have another question. Has the KMS to be set up in the same machine that runs the test?

The test works fine when I run it in one machine which has the KMS set up. On the flip side, when I uncomment the property kmsUri as follows:

  "test": {

         "fake": {

           "kmsUri": "ws://<kms_ip>:8888/kurento",

            "wr": {

               //absolute path on the KMS's machine or http resource

               //"files": "/tmp",

               //readable by the user that runs KMS (default is nobody:nogroup)

               //"filenames": []

            }

         },

And set it to a remote machine the test does not work anymore.

Regards,

Yago

[Radu Tom Vlad]

Hi, I'm pretty sure this was working before. Are there any errors when running the test(s)? 

PS make sure the file that the FakeParticipant is loading in the PlayerEndpoint is located on the same machine as the new media server and that the KMS process can read it.

[ya...@streaminggalicia.com]

Thank you! You pointed me in the rigth way. It was a permission issue.

Regards,

Yago

[Hemanshu Belani]

So what was the max number of participants supported??

[Ivan Gracia]

That depends on the server you are using to host your KMS instance. The bigger the server, the more users it will support.

Ivan Gracia

--

[Hemanshu Belani]

Thanks Ivan,

For testing purpose I'm using AWS t2.micro having 1 GB RAM & single core, but i'm still not able to figure out the maximum number of participants per call.

I tested with 6 people.

[Ivan Gracia]

I don't think you'll be able to go much higher. Give it a try! It depends a lot on quality of video, too.

Ivan Gracia

[vijay krishna]

Hi Ivan,

I am using kurento-room for conference calls.Well the issue is when i have 6-7 users i am getting request timeout on browser end for messages(publishVideo,receiveVideo) .Sure, Changing requestTimeout from 15 seconds to 20-30 seconds solves my problem.But it is really awful to wait for a response for 30 seconds and this issue isn't because of server load or browser load.It appears that my messages(publishVideo/receiveVideo) are taking 12-15 seconds just to reach the kurento-room client (I checked by adding logs) while those messages are sent to jsonrpcclient (js) right away.I am assuming it isn't entirely due to browser load . I am assuming jsonrpcclient(js) is holding(queueing) messages before sending to client(backend). 

Can you shed some light on what is exactly happening here and how i can optimise it .

[chapin]

I suggest you try this project： 

kurento-benchmark

The original is: 

  webrtc-benchmark

在 2016年3月22日星期二 UTC+8上午12:01:05，ya...@streaminggalicia.com写道：