AnsweredAssumed Answered

Recommended Architecture for Redundant PIEMDVB Interfaces

Question asked by jeff_denz on Feb 22, 2016
Latest reply on Feb 22, 2016 by jeff_denz

We are preparing to install PIEMDVB interfaces to replace our existing EVT interfaces.  The EVT interfaces are processing the batch data from Emerson DeltaV.  With EVT interfaces, redundancy was not possible other then just having a standby interface installed on another machine.  According to the PIEMVB interface manual, we can now use redundant interfaces.

 

We have 2 situations, in one case we can use the PIEMDVB interface to connect to the Batch Journal (SQL Server) and in a different case we have no choice but to process the EVT files.  I have already tested the redundancy when using SQL Server but I have questions about what happens when you process the EVT files.

 

Current Setup to process EVT files (non-redundant):

Currently we use the EVTSYNC application to copy the EVT files from the source to the interface node where the EVT interface is installed.  The interface then processes the files from that node.

 

Current Setup:

Machine A = DeltaV Data Source (EVT Files)

Machine B = PI Interface Node + Copied EVT files

Future Setup:

Machine A = DeltaV Data Source (EVT Files)

Machine B = PI Interface Node + Copied EVT files

Machine C = PI Interface Node + Copied EVT files

 

If we want to move to EMDVB interfaces and take advantage of redundancy, can we copy the EVT files to machine A and machine B and run the interfaces in a redundant mode?  If Machine B is the active node, does Machine C still process EVT files while it's running in backup or will it wait until it becomes active?  My concern is that if Machine C is not active for several months and IS NOT processing EVT files, when it becomes active it will attempt to process months worth of EVT files since they just continued getting copied by the EVTSYNC application but not processed.  In that case it could take hours for the interface to catch up and rename the .EVT files to .999 (meaning they have been processed).

 

Is there a better recommended setup is we want redundancy for EVT processing?

Alternatives I can think of: 

1)  Processing the EVT files on Machine A (I don't want to do this, I don't want to touch the source files at all)

2)  Copying the files to a different node and allowing both machine B and C to process the same files (seems like too much wasted overhead to bring a 3rd machine)

 

Has anyone attempted to get such redundancy working and can you share some lessons learned or does anyone at OSI have more details on how EVT redundancy should be setup?

Outcomes