3CX SBC in V18

[Chris Walters]

Have there been any new stress tests of the 3CX SBC in V18? Just curious how accurate this information is after 2 years: https://www.3cx.com/blog/releases/sbc-performance/

Was thinking of using one of the Pi4 B+ with 8GB Ram and 32GB SSD from Canakit. Really what I'm concerned about is concurrent calls and total BLF allotment. Thanks!

 

[VasilisV_3CX]

Hi @Chris Walters,

How large is the license you're using?

The stats you've provided should cover most scenarios with relative ease.

 

[Chris Walters]

One cloud server on a 48SCEnt system with 7 offices, largest site has 75 users, 140 total users.

 

[Evolute IT]

One cloud server on a 48SCEnt system with 7 offices, largest site has 75 users, 140 total users.

Put a SBC per-site, split the phones to each of them. You'll be fine.

One Pi can handle 40-50 phones easily.

 

[VasilisV_3CX]

Hi @Chris Walters,

Refer to the hardware guide here: https://www.3cx.com/docs/recommended-hardware-specifications-for-3cx/

 

[Chris Walters]

Hi @Chris Walters,

Refer to the hardware guide here: https://www.3cx.com/docs/recommended-hardware-specifications-for-3cx/

My concern is the BLF count. They are going with Fanvil X6U phones for the higher BLF count. I don't want to sell them phones capable of 60 BLFs and then have them only able to utilize 10 per phone.

If I am reading that guide correctly, each Pi4 SBC could support up to 500 BLFs (50ext x 10BLF). From other threads I have read, this is more an upper limit that can be divided across handsets in many ways. If I understand that correctly, I would need 1 SBC per 8 phones to cap out at 480 total BLFs?

Is it common to have multiple SBCs at one location? My previous cloud deployments have been small enough that I just create DHCP reservations for the handsets and then create SNAT rules for the phones on the firewall.

 

[Evolute IT]

My concern is the BLF count. They are going with Fanvil X6U phones for the higher BLF count. I don't want to sell them phones capable of 60 BLFs and then have them only able to utilize 10 per phone.

If I am reading that guide correctly, each Pi4 SBC could support up to 500 BLFs (50ext x 10BLF). From other threads I have read, this is more an upper limit that can be divided across handsets in many ways. If I understand that correctly, I would need 1 SBC per 8 phones to cap out at 480 total BLFs?

Is it common to have multiple SBCs at one location? My previous cloud deployments have been small enough that I just create DHCP reservations for the handsets and then create SNAT rules for the phones on the firewall.

Yes you can have multiple SBCs per site and split phones across.

However, you might wanna check for a more powerful Intel/AMD PC and run the SBC from there. This way, the cap is way higher than the Pi.

 

[Chris Walters]

Yes you can have multiple SBCs per site and split phones across.

However, you might wanna check for a more powerful Intel/AMD PC and run the SBC from there. This way, the cap is way higher than the Pi.

@VasilisV_3CX how many BLFs could a debian SBC running on a Ryzen 5 3600 with 16GB of RAM support?

 

[VasilisV_3CX]

Hi @Chris Walters,

The 100 is a soft limit, you could get more working on there but I can't say at what level it'll fail.

 

[Evolute IT]

@VasilisV_3CX how many BLFs could a debian SBC running on a Ryzen 5 3600 with 16GB of RAM support?

We run a i3/4GB/SSD for 80 phones with around 20-30 BLF per phone.

No issues.

 

[Chris Walters]

We run a i3/4GB/SSD for 80 phones with around 20-30 BLF per phone.

No issues.

I guess maybe I'll try an i5 nuc with 8GB ram and SSD and see if I can get 75 phones with 50 blf. If it doesn't work, I'll build a small ryzen 5 3600 box with 16GB of RAM and it sounds like that should handle it no problem.

 

[BrenttG]

@VasilisV_3CX how many BLFs could a debian SBC running on a Ryzen 5 3600 with 16GB of RAM support?

An i5 and 16 gb ram is a bit much, we have Atom E3845 Quad Core boxes with 4GB ram running 100 phones in several cases, with 8-10 BLFs per phone. We use the Atom E3845, and the i3 7100U boxes with amazing success as both SBCs and Firewalls.

The key is do not get one of those cheap little boxes with a low end Celeron, they mislead on specs, Celerons have no CPU cache, and are like trying to ride a bike through pudding. the Celeron J1800 and J1900 CPUs are especially bad, and actually have keyboard lag they are so slow, imagine that handling your audio channels, XD

Take one of these, put on 4GB RAM, 8GB if you really want to, and a 32 GB SSD, and your done. We use them, they boot up in LITERALLY 30 seconds or less with the 3CX SBC installed, and they support always on, meaning no one has to go press the power button after a power outage, if it has power it will turn itself on, just make sure that is enabled in its BIOS. Auto Power on: enabled, don't set it to last state, it needs to be set to on.
https://protectli.com/product/fw6b/

IT also has a nice massive heatsink on top, and no need for cooling fans that might fail after a few years anyway.

We order these in bulk, 50 at a time, because we use them for pfSense Firewalls as well, we have so far ordered and deployed like 4-5 batches, and only had 4 that were bad(one was dropped), so that's like 1 or less per 100 being faulty.

 

[Chris Walters]

An i5 and 16 gb ram is a bit much, we have Atom E3845 Quad Core boxes with 4GB ram running 100 phones in several cases, with 8-10 BLFs per phone.

The key is do not get one of those cheap little boxes with a low end Celeron, they mislead on specs, Celerons have no CPU cache, and are like trying to ride a bike through pudding.

Take one of these, put on 4GB RAM, 8GB if you really want to, and a 32 GB SSD, and your done. We use them, they boot up in LITERALLY 30 seconds or less with the 3CX SBC installed, and they support always on, meaning no one has to go press the power button after a power outage, if it has power it will turn itself on, just make sure that is enabled in its BIOS. Auto Power on: enabled, don't set it to last state, it needs to be set to on.
https://protectli.com/product/fw6b/

IT also has a nice massive heatsink on top, and no need for cooling fans that might fail after a few years anyway.

We order these in bulk, 50 at a time, because we use them for pfSense Firewalls as well, we have so far ordered and deployed like 4-5 batches, and only had 4 that were bad(one was dropped), so that's like 1 or less per 100 being faulty.

My concern is the BLF count. You're still talking about 800-1000 BLFs in your scenario. I'm talking about up to 4500 (75x60) at one location.

 

[BrenttG]

I understand that, but on the SBCs that we use, The CPU usage i see, is more less un-measurable.

If you need the i5 just to ensure there is never a slowdown or an issue, here are a couple more very cost effective options, two of which will shoot you to the moon CPU power and RAM Wise

i5 Dual Core @ 3.1 GHz, Hyperthreaded, 64 GB RAM max, Support for 2 SSDs - Flying High
https://protectli.com/product/fw6c/

i5 Quad Core @ 3.4 GHz, Hyperthreaded, 64 GB DDR4 Max, Support for 2 SSDs - Breaking High Orbit now
https://protectli.com/product/fw6d/

i7 Quad Core, Turbo up to 4.0 GHz, Hyperthreaded, 64 GB DDR4 Max, Support for 2 SSDs - an actual moon shot
https://protectli.com/product/fw6e/

You can go as high as you would like to with these devices, but honestly, I have never seen and SBCs usage broach the first item on the list above.

 

[Chris Walters]

I understand that, but on the SBCs that we use, The CPU usage i see, is more less un-measurable.

If you need the i5 just to ensure there is never a slowdown or an issue, here are a couple more very cost effective options, two of which will shoot you to the moon CPU power and RAM Wise

i5 Dual Core @ 3.1 GHz, Hyperthreaded, 64 GB RAM max, Support for 2 SSDs - Flying High
https://protectli.com/product/fw6c/

i5 Quad Core @ 3.4 GHz, Hyperthreaded, 64 GB DDR4 Max, Support for 2 SSDs - Breaking High Orbit now
https://protectli.com/product/fw6d/

i7 Quad Core, Turbo up to 4.0 GHz, Hyperthreaded, 64 GB DDR4 Max, Support for 2 SSDs - an actual moon shot
https://protectli.com/product/fw6e/

You can go as high as you would like to with these devices, but honestly, I have never seen and SBCs usage broach the first item on the list above.

Thanks for the suggestions!

My biggest overall concern is what 3CX lists as "recommended" so that if I run into issues, I know that I'm within the range of what they consider supported for the hardware. The last thing I want to do is open a ticket with 3CX and be told "that's not supported hardware".

I was really hoping @VasilisV_3CX could point me to some resources for supported SBC hardware for larger deployments. I have my eyes set on some potential clients that would be up to 10K BLF buttons at one site. It would be most cost effective to have 1 system capable of handling that traffic.

 

[BrenttG]

@VasilisV_3CX @ChrisC_3CX @AndreasY_3CX Guys have any thoughts on the above question? Would he be considered ok for support with the hardware i mentioned above?

 

[SweetAction]

Just a couple of thoughts (I too want to know what 3CX says for deployments with this many phones and BLFs though)

-You could run multiple SBCs per site (not 3CX HA but your own with your own Load Balancer and using DNS hostnames, round robin, etc)
-3CX SBC HA is active/passive but I would recommend anyways for deployment this large if you don't do your own HA
-In fact I would take a decent machine and install <virt platform of choice> then a couple SBCs on the virt platform. Doesn't stop physical hardware issues though
-Really at this size sometimes it make sense to just do a Site2Site tunnel

 

[BrenttG]

Just a couple of thoughts (I too want to know what 3CX says for deployments with this many phones and BLFs though)

-You could run multiple SBCs per site (not 3CX HA but your own with your own Load Balancer and using DNS hostnames, round robin, etc)
-3CX SBC HA is active/passive but I would recommend anyways for deployment this large if you don't do your own HA
-In fact I would take a decent machine and install <virt platform of choice> then a couple SBCs on the virt platform. Doesn't stop physical hardware issues though
-Really at this size sometimes it make sense to just do a Site2Site tunnel

For big companies like this they usually have a VMware ESXI, or HyperV or something on premise, we have a number of clients where we installed the SBC as a VM in their Hypervisor, namely, we have a 911 Center, and their's is in their VMware Cluster that also runs their windows servers, so it has HA, failover, etc, via VMware. But most situations don't have that option.

 

[NickD_3CX]

Thanks for the suggestions!

My biggest overall concern is what 3CX lists as "recommended" so that if I run into issues, I know that I'm within the range of what they consider supported for the hardware. The last thing I want to do is open a ticket with 3CX and be told "that's not supported hardware".

I was really hoping @VasilisV_3CX could point me to some resources for supported SBC hardware for larger deployments. I have my eyes set on some potential clients that would be up to 10K BLF buttons at one site. It would be most cost effective to have 1 system capable of handling that traffic.

Hi!

OK, let me help you interpret this here because I do understand that it can be a bit confusing sometimes:
https://www.3cx.com/docs/recommended-hardware-specifications-for-3cx/

First, to explain the limit.
When we say 100 IP Phones / 1000 BLFs (10 per device), in which case you have "2 upper limits", the general rule is, the limit is whatever you reach first.
This means, if each IP Phone would have 5 BLFs, this doesn't mean you can have 150 of these on a single SBC.
Reversely, if each IP Phone would have 20 BLFs each, you couldn't still have 100 IP Phones.

In your case (75 x 60), I would split it among 4-5 SBCs on the same premise if could.

Now onto the specs. You talked about the Ryzen 5 3600 (6 cores, 12 Threads/vCPUs).
For an SBC to handle up to 100 IP Phones/1000 BLFs, you need "4 vCPUs Intel i7 / 2 GB RAM / Debian-based SBC".
This Ryzen as far as clock speed goes would be equivalent to the Ryzen 5.

So, this CPU would be an overkill of a single SBC.

Here's an idea. By the time all are going to be on the same premise, get a slightly bigger CPU (say, 20 CPUs or more), turn it into an ESXi server with 32 GB RAM, then spin up 5 VMs, each one with 4 vCPUs / 2 GB RAM and each one can be its own SBC.
Added benefir, you can make use of some nice features like snapshots, backups, high availability, etc, that hypervisors have to offer.

Longer answer than I expected, but I hope I solved any questions.