Clients Slow To Login/logoff

[zerox22x]

it seems that sometimes our clients take a little bit of time to login or logoff aswell as doeing signalling, in some cases its just login and logoff but the slower pc seem to have issues when it comes to for example hanging up a call and less then a second later picking up a new one (nothing happends)

[Tom Wellige]

Do all your server and client machines comply with the hardware recommendations of Swyx?

 

What else is running on the server or the client machines?

 

There are a ton of possible reasons for such behavior. So somehow we have to start to pin it down...

 

 

You should run the windows performance monitor on the server machine to be able to see the cpu load and memory/handle usage of the server process in such moments.

[zerox22x]

all machines are at minimum an i3 at 2.5 Ghz, 

4gb ram

 

the server is a Xeon e5-2420 @ 1.9 Ghz with 46 gb ram

currenty is at 70% memory load and 10% cpu usage

in the hyper-v itself

memory load doesnt get over 75%

cpu tends to average on about 14% 

when actively monitoring it occasiannaly goes to 15% total use (12 virtual cores)

 

its mostly only during login that its slower thenwhat it used to be (2-3 seconds now 5-8 seconds) and sometimes when trying to transfer to another person or picking up a line after hanging it up (client seems to freeze for a few seconds)

[Tom Wellige]

So what other containers/software do you have running on your hyper-v server that generate temporarily cpu load?

[zerox22x]

an sbs2011 server hosting ad and a fileshare however when it happends all clients are already logged in

and a debian server hosting a database

also another test server however this one barely gets any load if it all

[Tom Wellige]

Well, something must slowing down your system as you wrote:

 

 and sometimes when trying to transfer to another person or picking up a line after hanging it up (client seems to freeze for a few seconds)

 

That's why I suggested to run the performance monitor all the time to be able to check the numbers afterwards.

 

Do your users also experience any stuttered announcements?

[Virikas]

the server is a Xeon e5-2420 @ 1.9 Ghz with 46 gb ram

currenty is at 70% memory load and 10% cpu usage

in the hyper-v itself

memory load doesnt get over 75%

cpu tends to average on about 14% 

when actively monitoring it occasiannaly goes to 15% total use (12 virtual cores)

As Swyxware isn't the only VM running on that host you should think about your container configuration.

The e5-2420 is a six core processor which has hyperthreading enabled so that you get to a total of 6 physical and 6 virtual cores. Microsoft counts multithreading cores exactly the same way they count physical cores. In my personal experience I can't agree to that. A hardware with 12 physical cores performs much better in a Hyper-V environment than a hardware with 6 hyperthreading enabled cores. Microsoft also states that a VP:LP ratio of 1:1 performs better than 1:2 or 2:1. In your case you have assigned all 12 (of which i personally would only count 6..) cores to your swyxware installation resulting in a correct 1:1 ratio. But you also have 4 other machines also concurring for cpu ressources. So you have a total of 4*12:12 = 4:1 VP:LP ratio which is not far away from the maximum tested and supported 6:1 scenario. You should lower your ratio closer to 1:1 especially when there are realtime application like Swyxware.

Microsoft provides detailed strategies for performance analysis on hyper-v environments you should do.

Besides that: Swyx itself is mainly working like a single threaded application, where most of the performance on each single comes from raw cpu throughput per core. Throughput per core can only (within the discussed environ ) be achieved by higher cpu freqencies. With the e5-2420 you are using a processor with a rather low frequency.

 

its mostly only during login that its slower thenwhat it used to be (2-3 seconds now 5-8 seconds) and sometimes when trying to transfer to another person or picking up a line after hanging it up (client seems to freeze for a few seconds)

This might(!) result from multiple containers "concurring for cpu ressources" and your rather high VP:LP ratio, which you should analyse in depth using the tools provided by microsoft.