Clocking/Timing: To Virtualize or Not to Virtualize

Hello!

I’ve been putting in physical machines running Asterisk for over 10 years. They’ve always been physical machines and have always had Sangoma PRI cards.

Back in the day I thought I remembered reading that Asterisk required a telephony card for accurate clocking/timing - in fact Sangoma used to sell a USB device for accurate clocking/timing for servers that didn’t have any telephony cards (systems were using a SIP trunk).

I’m assuming that Asterisk has more accurate clocking now and telephony hardware is no longer required (https://wiki.asterisk.org/wiki/display/AST/Timing+Interfaces). Am I correct in this assumption? Would the timing be accurate enough clocking for 200 hundred phones, 30 or 40 concurrent calls, and MoH?

And if accurate timing is achieved without any hardware can this same sized system (200 phones, 40 concurrent calls) work in a Microsoft Server 2012 R2/2016/2019 Hyper-V Virtualized environment (assuming fast disk subsystem, 8 or 16 GB of RAM, and multiple CPU cores)? This question gets to the accuracy of the VM clocking specifically for Microsoft Hyper-V.

Any thoughts would certainly be appreciated!
Thanks! Dave

Yes, you are.

1 Like

Thanks, Dave, for answering my question about Asterisk timing interfaced.

Somehow my final question about whether a Hyper-V environment would be suitable for a virtualized FreePBX system wasn’t displayed correctly, but now I “fixed” it. Can anyone address this question for me?

Thanks for the help!

I know you are asking about Hyper-V, so this is not quite on target, but I run FreePBX on VMWare and it works great, I would not run it any other way.
…For what it’s worth

There are lots of virtualized platforms running FreePBX out there. Check back through the archives for lots of posts about running your system on other people’s hardware you have no control over.

We use hyperv primarily for pbx. Works fine. Most systems run around 50 simultaneous calls. No issues. Running hyperv in a cluster, we can live failover pbx guests to different hosts, no issues with calls

Dickson, thanks so much for your input - this was what I was looking for! I have customers with Hyper-V Clusters that I manage so I know I have enough CPU, disk, and RAM. I just wanted to know about the FreePBX clocking/timing issues with a large call volume. If you have 50 simultaneous calls, I’m curious to know how many extensions do you have on these systems? I’m dealing with customers who have between 100 and 200 at most so my guess is that you probably have more than that. Also, what version of Hyper-V are you using? Thanks again!
dave

Most are running between 50-200 extensions per guest, there is another that on here that runs closer to 300 users, but their call volumes are less than 20 usually0. Registrations aren’t a problem, its simultaneous calls that really drive CPU usage in my experience. I generally avoid any codec but ulaw/g711 to keep things in check.

If there are some HV specific numbers you want to see, let me know. Our environments are based Windows Server 2012R2 on Dell r710s (X5550) and r720s (X-E5-2670) connected to iSCSI equalogics. No issues moving images between the different hardware either. Usage is low at the moment, but it’ll ramp as the day progresses.

Thanks so much for your help! I appreciate it!
dave

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.