On Fri, Jul 6, 2012 at 7:27 PM, Stanislav Sinyagin <ssinyagin@yahoo.com>wrote:
OpenVZ and XEN VPS'es are usually fine for quick prototyping and feature tests. Also they would be fine for private PBXes with low call volumes. For hundreds of simultaneous calls, it's probably cheaper to rent dedicated hardware servers. On the other side, if you have some smart load balancing, probably you can build a scalable design with VPS'es and eventually migrate to physical servers when volumes are high enough.
also for a calling card business, probably the most critical CPU resource would be needed for the IVR and real-time billing, but not for the calls themselves.
Thanks Anthony and Stanislav. I think I have enough warning along with some encouraging words to build a strategy. Thanks again., Now the search for the right IaaS provider begins.
------------------------------ *From:* A E G <all.eforums@gmail.com> *To:* FreeSWITCH Users Help <freeswitch-users@lists.freeswitch.org> *Sent:* Saturday, July 7, 2012 12:23 AM *Subject:* Re: [Freeswitch-users] RTP delay > 100ms by performance testing with freeswitch as trancoding device SILK to G711a
On Fri, Jul 6, 2012 at 5:43 PM, Anthony Minessale < anthony.minessale@gmail.com> wrote:
I would be insane to endorse using it in a virtual world. The support expectations would consume the rest of my life. I try to make the code work well on real boxes and if it works on virtual ones then that's a bonus =D
We have good luck with openVZ since its a single real kernel and virtual runtimes but either way we do not recommend it so its a use at your own risk kind of thing.
Yup agree and acknowledge that you wouldn't endorse or offer to support it, and nor should you. Simply trying to pick your (and anyone else who wants to chime in) brain on the topic. Totally PoC'ing it at my own risk :)
So, in that spirit, just one last message about it. Just like sharing the experience with OpenVZ, let's just say that I am insane and I still want to try it in a private cloud. Would you say the resource consumption of FS (assuming it's just receiving the call, having it routed by a custom routing engine via ESL, and switching it) be more than that in the "real" world for the same call volume barring any clock sync/skew/jitter issues? Would something like Cloudstack on hardware (like the one you recommended earlier) with a FS VM with 12 vcpu cores, 24GB RAM and DAS be able to handle something like 300-400 concurrent calls or am I out of my mind?
Thx
On Fri, Jul 6, 2012 at 3:59 PM, A E G <all.eforums@gmail.com> wrote:
Yup got that, and are you saying the same is applicable and true even if one was to try and run FreeSWITCH on a private cloud or in a virtual environment? assuming of course that those tips would then apply to the machine on which the Hypervisor will run.
Just trying to get a handle on whether running FreeSWITCH to do something like wholesale or calling card traffic in a purely virtual environment is / has proven to work. You probably know the most in terms of all different environments people are running FS in, and if you (or they) have pointers specific to it being run in a virtual environment.
Thanks so much
On Fri, Jul 6, 2012 at 12:54 PM, Anthony Minessale <anthony.minessale@gmail.com> wrote:
Like I said "Really Really Nice Motherboard", as many CPU as possible/affordable, and a good chunk of RAM. Motherboard is the most important, a cheap motherboard with a ton of cores is a waste.