Scalability HUB/Networks

App_rpt-users
1

Hubs/networks Scalability.

In the ASL software we have grown a network to have 60-70 nodes connected. We are reaching a limit where we can no longer connect to other hubs of a similar size.

We have also tried scaling on the new beta ASL but for some reason when we connect to other similar size hubs it will crash the node. Is there any plans in refactoring app_rpt code to improve scaling past 70-80 nodes and to develop a load balancing mechanism to stop the node from crashing after a significant amount connected? I know that HAMVoIP have worked on this issue and have developed a HUB version of the software with improved performance for networks scalability. We believe it’s somthing to do with the exponential time increase in the data structures used, but this is way out of my league. We are stretching the very fabric of asterisk at best!

Please can you shed some light on this issue as since we have launched our new network we are experiencing these crashes now.

Thanks
Shane
M0VUB

2

It’s definitely something we can add in the future. Maybe an option of hub=1 to reduce IAX messaging and statpost keyups.

The new beta puts coredumps in /tmp. If you still have the beta running, can you send over the /tmp/core file to me at rob@allstarlink.org?

Cheers,
Rob

3

Hi Rob. Yes sure I’ll get Bruce to pull the core dump file. We are running 2 beta servers now so we are more than happy to test to destruction and send what ever you need.

Thanks
Shane - M0VUB