So i would like to see how insane I am before venturing too far down this road. We are a small CSP with about 30 customers. Most have unique requirements so we run each of those FPBX 14/Asterisk16 in a 3rd party cloud IAAS provider (Vultr). They work well until there is a noisy neighbor or a switch goes sideways.
My thought is to do as follows:
A. Secure a 1/2 rack at a very good Data Center
B. Put (2) HyperV 2016 servers with 256G + storage in a cluster (for failover); connected to a third 2016 Server running SMB3 and fiber connected to the 2 custered servers (RAID 10) - cluster is run using the recommended 3-4 NICs for data isolation.
C. Run all clients in the two servers - and as we get beyond 50 or so VMs (memory dependent) add another of the same style server to the cluster (n+1)
D. We don’t intend on running the FPBX HA module, but leverage HyperVs tools + shared SMB3 storage
E. PFsense in front of all of it w/ all Virtual IPs
Is this crazy for 100+ VMs w/ FPBX14 / Asterisk16?
Boy would give a leg for a Multi-Tenant FPBX version as you can imagine!
I think part of it depends on how much time you want to dedicate to administering the infrastructure.
It always seems like a good idea when you start on a road like this, but you soon find out that maintaining hosted infrastructure is its own problem domain in and of itself and there are a number of problems you probably will encounter that they are dealing with that you’re not aware of up front.
Weird symptoms (strange slowness or customer performance problems, etc) that manifest themselves as telephony or other systems problems but are actually rooted in something entirely hard to connect (sometimes disk I/O or network backplane problems) and oftentimes challenging to diagnose unless you’re already an Asterisk expert (at sometimes a developer level).
