Hi our club currently has asl3 hosted on a cloud server, i was just looking to get info on making jt dedicated hardware at the main repeater site. Is there anything specific i should know about running it on dedicated hardware? Or is it just an easy debian install and run asl3? I know the raspberry pi’s can only handle 30ish connections and we are wanting something that can handle more between allstar and echolink. Thanks in advance
I think this is ancient lore. A Raspberry Pi 5 is a very capable system that should be just fine for what you’re wanting. It’ll be more powerful than a lot of the old “micro” PCs or “thin client” PCs.
Yep. We’re in the process of setting up a three receiver voting system around a Raspberry Pi 5 at a repeater site, with another Raspberry Pi handling connections from the outside world to keep things buffered and isolated a bit.
The rPi5 will handle the direct connection to the repeater, the RTCM server and Supermon (they prefer Supermon over Allmon, and have done a lot of customization to it). The public connections from ASL/Echolink will be on it’s own rPi. So, in theory, if something goes really wrong, it won’t stop the repeater from repeating. Probably a good idea to separate connections from the outside world from whatever is directly connected to the repeater. After all, your USB channel drivers use clock cycles that take away from other resources when in use
An interesting point about connections though:
Back in September, I ran a test using ASL3 on a Zeon 8 core workstation with 64GB RAM on a multi-Gigabit connection. Entry level as far as Zeons go, but plenty more than ASL should ever use. What I discovered was that, on a single node, things got really rough at around the 85 connection mark. I was expecting it to not bomb out so early. At least it didn’t crash. It might have done better with some load balancing between nodes internally, which would require quite the dial plan modification.
I have done this internal load balancing thing on HamVoIP before, and managed about 120 connections on a single Raspberry Pi 4 before things got really choppy, but it was crying a little bit at that point. I’ve not yet tried to replicate that with ASL3, because a lot of the dialplan syntax has changed in modern Asterisk, and I just haven’t had the time to study it.
The basic premise is you have a few nodes, and each successive connection goes to the next node.
So, let’s say you have a node, 508420, which accepts connections. The first to connect gets routed to a private node, let’s say 1900. The next connection goes to 1901, the next to 1902, and the next to 1903. Now, the fifth connection connects to 1900, sixth to 1901, and so on. Then, all the nodes are connected to 508420. The theory is that each individual internal node handling less connections will increase the over all stability at higher connection counts, and, at least with HamVoIP and a bunch of dialplan changes, that does seem to be the case, but I don’t know if they did anything to app_rpt to help that along. It isn’t open source, after all.
I think you will find stream limitation not per node, but per server.
You have many potential bottlenecks with high stream counts.
CPU, Memory, Internet Bandwidth.
All depending on the resources at hand and the processes running at any given moment.
The only thing I will really add to this is that if you are or fear near limits of some or any of this, turn Allison OFF. And don’t announce anything except the command at hand locally.
When you have a lot of connections and you are allowing connection announcements of foreign nodes, that can be overwhelming and stack up in buffer until played or it crashes.
Better schemes can be found by using external servers for some connections.
Like using a hub server node to take 25+ connections with one route to the other system.
There is more overhead in 2 nodes with 25 connections each than one with 50 on the same server as measured in cpu and memory…
Of course that test was done 3 versions of software ago. May be different.
The WAN System used one server for personal nodes and one other for repeaters.
But I don’t know their current scheme. They at least had plenty of inet bandwidth and might have over 100 nodes at any given time. (not anymore).
Just something to think about in planning for this stuff.
But I will say that I probably still hold the record for most connected network stations back in 2014 when we created the ‘Lincoln highway link-up’ for those that remember that.
And those folks over on the WAN system were instrumental in networking everything east of Ohio.
There are lessons to be learned, but most of us really learn them by trial and error.
I encourage trial and error. You will not make it better if you don’t break it once and examine the pieces.
We already have one pi5 as the main node connecting into the repeater, So we just need another to be the host for allstar and echolink to replace the webhosted vm?
Unless you have other reasons to do it, there’s likely no problem putting it all onto the Pi5 already connected to the repeater.
I have a question i have a raspberry pi 5 but i dont no how to work it i try to run a image on it but its not worked at all do anyone knows about it