I posted the following on the Replicator forum and Rainer Herrmann mentioned that for some questions the Reptor is not the right forum.
Based on very slow replication (SLOG piled up - 20 replications per second - unstable Reptor) of mass data from Adabas on the mainframe to Adabas on Windows, I would like to know the following:
Is the Adabas Event Replicator/WCP latency sensitive and what are acceptable latency thresholds to keep the application from backlogging significantly enough to cause the mainframe process to become unstable?
Does the Replicator/WCP employ any slow start methodology or other way to optimize network performance?
How does the Replicator/WCP handle lost data within the replication stream or lost/retransmitted TCP/UDP packets?
Considering the Replicator/WCP, can the data sent between the host and client be safely encrypted with IPSec?
Do you have other customers successfully using the Replicator/WCP over an IPSec VPN tunnel?
Replicator ARF 3.2.1
Rainer Herrmann’s response:
The event replicator is designed not to impact the mainframe source process.
If the target can not keep up with the load, replication can be written to SLOG instead and the assumption is that this should be faster than the source can create new replication data.
However, there have been enhancements and important corrections in ARF321, which addresses problems in this area and I recommend that you move to this version
The questions you raise with regard to Net-work issues are to the wrong product
Reptor uses WCP like any other Adabas client
Currently the reptor does not do special optimizations
In future it is likely that there will be optimizations like multicalls etc.
Lost data is handled by the Net-work protocol, not by the client (reptor)
Encryption (e.g. SSL in Version 6.2) must again be supported by the Net-work version not by the client.
So check the version of your net-work documentation what is supported.