1 Topic by Sam_China

  • Sam_China
  • AA China Chief Engineer
  • Offline
  • From: Beijing
  • Registered: 2007-09-28
  • Posts: 11

Topic: How to do a redundancy NION system

In Asia, i had some projects using redundancy NION system, some of them are 1 to 1 backup,the others are 1 to more. the technique as follows:
  To make a complete automatic backup system with NION, we need work by two steps: the first is auto-changeover to redundancy NION when the primary NION is down, including power failure, DSP stoppage, fan temperature deviant,LAN port communication stoppage and other you think turn to backup NION in need. the second step is changeover the CM-1 port to backup NION's this part needs another program to switch.

  1\NION's changeover: put a Project Link device with Blinker(act heartbeat) and CM-1 status and fan temperture meter etc. on the primary NION side, and then put a script on the redundancy NION side to judge whether to trigger changeover action, if yes the Script will be switch bundle number from 0 to valid number

  2\CM1's changeover: using a third party emulated Nware program with a SNMP module python script to monitor the redundancy NIONODE status once it starts running the script will be sending a command to change parameter of Request_Bundle_Priority of redundancy CM1 from default 0x1010 to higher. note that Request_Bundle_Priority  is different with Priority in NWare CM1 advanced configurable table. please refer to CM1 Programmer's Reference manual P103 about txPriority description.

NOTE: for redundancy design, all NIONODE are pulled out the analog I/O card/s, if you want to use them be careful of impedance of input card and relay security of output card.

about 1 to more redundancy design, if someone needs it I'll post other topic.

2 Reply by Sam_China

  • Sam_China
  • AA China Chief Engineer
  • Offline
  • From: Beijing
  • Registered: 2007-09-28
  • Posts: 11

Re: How to do a redundancy NION system

Actually we can through pandebug.exe to change the parameter of Request_Bundle_Priority firsthand, but this way couldn't operate automatically ,You must manual reenter one more after rebootting the redundancy NIONODE.

3 Reply by cwa

  • cwa
  • cwa
  • Peavey Technical Director
  • Offline
  • From: Boulder, Colorado USA
  • Registered: 2007-08-13
  • Posts: 193

Re: How to do a redundancy NION system

Good thoughts on the topic.  Building fault tolerant systems is complex and it sounds like you've built a few.  I am curious about your approach to 1 to many.  Do you have a way to do this automatically?

4 Reply by Sam_China

  • Sam_China
  • AA China Chief Engineer
  • Offline
  • From: Beijing
  • Registered: 2007-09-28
  • Posts: 11

Re: How to do a redundancy NION system

Yes, I had built a few projects that's working on 1 NION to more to do redundancy

automatically in China this year and all is working well. I'd like to share this solution with

everybody.
For 1 to more redundancy design we need to bethink of two states: with XDAB and no XDAB.

Underside example is 1(for short C) to 2(for short A and B) to do a backup.
NOTE: for any state we need put a third party PC based computer it's running a emulated

NWare++(for short E) with a PYSNMP module group library. This E has 2 intentions, 1st  for monitor and manage the redundancy CM1 txPriority parameter ,2nd for monitor which NioNode is broken and launch a matching backup file to redundancy NioNode.
No XDAB state:
   1/ Copy the NPA file from A and then create a new file rename it and its inside ProjectLink

server  from mainA.npa and main_A_status to backupA.npa and backup_A_status.
   2/Create a shortcut icon for the backupA.npa it's named backupA.lnk and put it under the

directory of d:\backup_demo.
   3/Repeat 1st and 2nd steps to create the backupB.lnk to same folder.
   4/Deploying the backupA.npa and toggle this role with NioNode C.
   5/Deploying the backupB.npa and toggle this role with NioNode C again.
*  6/Login the web status to erase the role to NioNode C and then it's in standby mode
   7/Here, the emulated NWaer E to monitor the ProjectLink status to both A and B ,suppose

that you unplug the RJ45 from A LAN port then, the main_A_status LED will be die on the E.

so this action can be launch the backupA.npa to C by a python command the code as follows:

import os


if inputs[0].string_get()=='true':
  
   os.system(d:\\backup_demo\backupA.lnk)

this action can be reloading and deploying the backupA.npa to NioNode C automatically.
   8/When the backupA.npa deployed ,it will be sending the named backup_A_status

ProjectLink to E and then this lighted LED will be trigger a SNMP command to CM1 of

NioNode C to change its parameter of Request_Bundle_Priority to highest for looting the

Bundles Control from NioNode A as topside saying.
* if you understand all then you can bypass the step 6.

5 Reply by Sam_China

  • Sam_China
  • AA China Chief Engineer
  • Offline
  • From: Beijing
  • Registered: 2007-09-28
  • Posts: 11

Re: How to do a redundancy NION system

With XDAB state:
   Because in a interlink system with XDAB we just need to program ONE role npa file,so that means we need to provide multiple backup npa file for whichever dead NioNode. in other words if the looped system includes 8 NioNodes + 1 backup, so you need to provide a main program plus 8 backup programs.
   We just utilize the XDAB's loop redundancy speciality to do the backup solution,the way is same with above no XDAB state but here have some especial problems:
   1/One loop of XDAB only can including one redundancy NioNode.
   2/Once one NioNode was broken the global NIONs with XDAB will be muted to waiting for the reloading NPA file for all NIONs.
   3/In case of more than one NioNode was broken, may be the interlink system with XDAB will be breakdown that's very dangerous.

Though this solution is very cheap than 1 to 1 backup but that has two deathful problems: the first is the reloading program time will be take a long time that up to your design may be 10 mins or 30 or 50 mins, can you put up with it ? the second problem is above 3/
A building project is the reconstructive PA system for HongKong int. airport  this year,I'll put 9+1 NioNodes in the ControlMatrix system.

Why not 1 to 1?  I don't know^^^^^^^^^^^^money?^^^^^^^^^^may be^^^^^^^^^^