You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@guacamole.apache.org by "Kurt.Ding" <we...@fastonetech.com> on 2022/03/07 10:26:16 UTC

Re: It is a efficient implementation about tunnel in guacamole common project that uses nio

Hi Gucamole Team

Could you make some suggestions about  tools  which   is used to run  
vnc/rdp  benchmarks?

Thanks

Kurt


在 2022/2/16 16:07, Kurt.Ding 写道:
> Hi Mike Jumper
>
> I need some days to do benchmarks . After that I will provide you our 
> test report.
>
> I am looking forward to hearing from you.
>
> Thanks Kurt
>
> 在 2022/2/16 15:14, Mike Jumper 写道:
>> Can you provide any specific benchmarks/metrics showing that there is an
>> appreciable amount of CPU time being spent in thread context switches 
>> prior
>> to the proposed change, and a corresponding improvement after the 
>> change is
>> in place (for the same load)?
>>
>> - Mike
>>
>>
>> On Tue, Feb 15, 2022, 22:44 Kurt.Ding <we...@fastonetech.com> 
>> wrote:
>>
>>> Hi Nick
>>>
>>> I'm glad to hear from you. Let me answer your questions and then we
>>> could have a discuss.
>>>
>>> Firstly let's assume one user  create hundreds remote connection, then
>>> in the linux server  which is deployed websocket  container the java
>>> process creates  hundreds of thread , it cause high cpu usage .
>>>
>>> In our production  server(8 core 16g)  ,  we meet  the case dozens of
>>> users create about 100 remote connections, unfortunately every user
>>> can't use vnc/ssh smoothly . Then we fix this problem with upgrade our
>>> server spec
>>>
>>> Secondly, as we all knows , many threads running in one server results
>>> in  thread context switch , and wouldn't increase the number of request
>>> per second.
>>>
>>> I hope this will make sense for you. And Your suggestion about use
>>> existing classes is a good idea. I will follow your idea and have a 
>>> try.
>>>
>>> I am looking forward to have a discuss with you.
>>>
>>> Thanks Kurt
>>>
>>> 在 2022/2/15 22:23, Nick Couchman 写道:
>>>> On Tue, Feb 15, 2022 at 2:11 AM Kurt.Ding 
>>>> <we...@fastonetech.com>
>>>> wrote:
>>>>
>>>>> Hi Guacamole Team
>>>>>
>>>>> Was I wrong about this  thread running model? Look forward to your
>>> reply.
>>>>>
>>>> This is a bit outside my area of expertise, but I guess I'd start by
>>> asking
>>>> the following questions:
>>>> * What actual problem are you trying to solve? What problem is thread
>>>> creation causing for you? Why is it that you want to limit or 
>>>> reduce the
>>>> number of threads created? Is reducing the number of threads actually
>>> going
>>>> to resolve whatever problem you're concerned about, or is it just 
>>>> going
>>> to
>>>> move the problem to another place?
>>>> * Rather than creating a WebSocket bridge, would it make more sense to
>>>> integrate your changes into one of the existing classes, perhaps 
>>>> either
>>>> extending an existing GuacamoleTunnel implementation or just
>>> implementing a
>>>> new GuacamoleTunnel that could then be used within the code? That 
>>>> would
>>>> seem to avoid this extra step of having to tunnel your WebSocket
>>>> connections through another server/class.
>>>>
>>>> -Nick
>>>>