Users are complaining about lag in audio resulting in people talking over each other.
Standard conference module (not Pro)
All modules current as of right now (2:12 PM PST, 01 Jan 14)
Digium G200 PRI interface to Comcast PRI
No external SIP connections or ports open; VoIP portion of system is internal only.
App Memory hovering around 34%
Proc gets into the 30s and 40s when 6-10 calls are running.
System is running in Hyper-V with 1 virtual proc and 2GB of RAM. Could easily bump this up if it would help.
Users are also reporting occasional dropped external calls.
Get off Hyper-V. Conference Rooms are tough in virtualization along with other things.
Due to cost, backup, and failover issues, that’s not a viable option. From a sales-cycle standpoint, not having to buy hardware other than the PRI interface is a huge deal.
Can you provide some detail about where the bottlenecks are?
Windows perhaps? There are better (and free) virtualization solutions, look at Proxmox and qemu if you are comfortable with linux, both run lean mean kernels that don’t need to kowtow to a greedy GUI based OS and present the client machine with close to hardware performance, with the right hardware (CPU) you can also easily present PCI/USB hardware on the host machine to the guest.
KVM is what we would recommend but conferences are real tough on any virtual platform. You will keep having issues when using conference rooms without real hardware and timing.
qemu is now pretty well KVM unless you choose XEN (a harder way), as I suggested, with vt-x and vt-d (intel) CPU support, you can easily present cheap chinese digium clones as a timing source on your pass-through PCI bus.
I would have to agree though that a real production server belongs always on real hardware, which in your case could easily be a $30 cast off from the Goodwill guys, get two for when the first one breaks
Make sure you put a new HD in it. Actually with the plethora of Atom based thin clients you can build super cool micro FreePBX systems. Most of them don’t come with large enough flash drives but those cleaver Chinese have all the popular micro drive form factors in 32G or 64G capacities. Slicker than horse snot.
The other issue is the ability to snapshot the system before major upgrades.
I guess what I was asking about before is a technical reason that conferences on VMs is an issue. The “because Windows sucks” type comments aren’t helpful.
The server on which this deployment runs is a Dell PowerEdge T610 with dual six-core Xeon procs, 32GB of RAM, and the FreePBX VM vhd sits on a 15k SAS hardware RAID5 array. This isn’t some cast-off box that we slapped a server OS onto. The host OS doesn’t do anything other than run Hyper-V.
Not trying to be a PITA here, I’m trying to learn something.
Technically if you virtualie your kernel, you need to have a rock solid kernel timer or hardware clock to synchronize the audio bridges, Hyper-V just doesn’t have that ability. So If your Hyper-V on a Power Edge T610 running FreePBX sucks at Conference and a couple of old Supermicro dual core atom’s running Proxmox (no vt-d) can run a multi way conference without a hic-cup, and further can migrate,snapshot and clone into iscsi, nfs or local drbd disks within the cluster with seamless failover (fencing) then there is perhaps something to my argument.
My bet is that if you threw away you Hyper-V and installed Proxmox (minimal effort, maximal gain) on the same hardware, you just wouldn’t have any problems running half a dozen asterisk conference servers on it concurrently, live snapshots and clones just are not a problem.
So I don’t say Windows sucks to be unhelpful, I just say it because it does, at least when it comes to virtualization, just ask anyone who has ever made those comparisons