SPECS INQUIRY

[Leckbush]

Hi I have a 2nd Vicibox install on other pc here which I plan to use for our new campaign. Is the system will handle our campaign?

6 Agents will Dial (Manual Dialing)
US Campaign

SPECS:
i3 2nd Gen 3.10GHz - 4CPUs
4GB DDR 1333MHz Non-ECC
500GB of HDD

Vicibox Installed is:
Vicibox8.1 2018
VERSION: 2.14-695a
BUILD: 181116-1133

[williamconley]

Number of agents is relevant, but hardly conclusive. Number of active channels, number of ringing channels, number of logged in agents, codec (g729?), recording calls ... all these have huge effect on load. Aside from that *everything* has at least a minimal effect on load.

That being said: As long as you're manual dialing and those 4 cpus are physical NOT logical, you should be fine.

Do note that the Vicibox version contains three segments (eg: 8.0.1) not two. Your example: 8.1 is missing the last segment. That last segment is usually an "installer bug fix" and is therefore very relevant.

[Leckbush]

Well I can assure that they will dial manual only NOT predictive. About the CPU those 4CPUs means 4Thread, in other words 2CORE (HyperThread) = 4THREAD = 4CPU Logical.

Login Agents is 6-7(Manual Dialiang)
Active Channels, idk this but if its the "Max Trunk" it's set to 96 with 100 "Max Calls per Second"
Recording is force on all calls they made.
No, we're not using g729.

Correction on Version: Vicibox 8.1.2 181002

[williamconley]

4CPUs means 4Thread, in other words 2CORE (HyperThread) = 4THREAD = 4CPU Logical

1) Not all 2 core cpus are capable of 4 threads. Only some. And many 4 core cpus are also incapable of hyperthreading. It can also be turned on or off even if it's a capable CPU. Which is why I asked.

2) When a system is heavily loaded, there is an argument that "fake" 4 threads will overload sooner than "true" 2 threads with 2 cores without hyperthreading. Virtual cores is marvelous on a desktop system when you have multiple windows doing multiple things that will "just go slower" while you sit there staring at your screen. Vicidial, on the other hand, requires every tick of the cpu (impossible with virtual cpus) and you *will* notice when the system has to wait under heavy load. Call quality, db response, etc, may suffer. If you experience instability, consider turning off hyperthreading (may help, may not, but worth a shot).

3) All that being said:

You're *probably* still ok with 2 cores (with or without hyperthreading). Not as likely as it would be with a true 4 core system (with or without hyperthreading), but with only 6 agents manually dialing you're probably still fine.

PS: By live channels, I refer to the most channels that may be live at any moment in time. This is not a system setting, this is use. Think of it this way: Your car's speed is NOT what's at the far end of the speedometer. It's where the needle is pointed right now. There may be a governor on it that stops you from exceeding a specific limit, but the speed you are going right now is still whatever the needle is pointed at. Thus "max channels" is not the same as "live" or "active" channels ringing or in call.

[Leckbush]

Okay thank you for enlighten me about that I will depend on what you said that it will still be fine with 6 agents manually dialing

So If I understand, 1 call is equal to 1 live channel right?

[williamconley]

Code: Select all

asterisk -rx "core show channels"

during live production. that will show channels.

PS: Welcome to the Party!

[Leckbush]

Okay, I just execute that command on our Main Server whose on operation right now (6 agents Manual Dialing on Main server)

Server said 29 active channels, 20 active calls, 8309 calls processed.

Is this gonna be the same on my second server?

[williamconley]

That list is based on the calls on the server right now. The list on the new server will be based on the calls on the new server when you run the command.

[Leckbush]

Thanks william, I will update if what I can observe when the 2nd server is on operation. Many thanks

[williamconley]

When it's actually running, you should be more interested in:

Code: Select all

uptime

Which will show the Average Server Load for the past 1,5,10 minutes. Since you're using 4 threads, that number should stay below 2.0 (half of 4) in all three spots. If it goes over 2.0, your system is working a bit hard. If it goes over 4.0, your system is beginning to overload. But until 2.0 average server load ... you just don't care. Until 4.0, you shouldn't panic. But when it reaches 4.0, you'll need to look at upgrading or adding a 2nd server. This may be MUCH farther down the road of course.

On a single-server system, this information is also available at the bottom of the Real Time Screen (so you can avoid logging in to the Linux CLI, much easier to keep an eye on it there).

[Leckbush]

Noted on that william I will try keep it below 2.0

[Leckbush]

Hi William, just want to ask.... I have million of DNC to uploaded on this 2nd server, will this affect the performance of server when we start dialing manually after all the DNC are uploaded?

[williamconley]

Hi William, just want to ask.... I have million of DNC to uploaded on this 2nd server, will this affect the performance of server when we start dialing manually after all the DNC are uploaded?

Depends on your hardware. This is something you should TEST before you get into production. A million DNC is difficult to explain, though. It would be worse with autodial.

[Leckbush]

Okay noted on. I have another thing here about this millions of DNC. I uploaded this DNC on our MAIN Server and we're dialing normally on manual with that DNC uploaded. But when we're doing predictive/ratio dialing, the system(asterisk seems to be working normally on ratio at first, but after the few minutes. Like 5 minutes, the system wont automatically dial the suppose it would be. Like if I set the dial method to 10 Calls per agent, sometimes the system only dial 1 or 3 for that agent. Do you think its from the list of DNC we uploaded which count more than 10 million leads? Before this, its actually doing fine though.

[williamconley]

Still don't know where you got a million entries for your DNC. I have doubts.

Check your slow query log. If it involves the DNC lookups (entirely possible), then you have your verdict.