Rockwell Automation FactoryTalk Historian SE 3.0 H2H Interface User Guide - Manual (Page 21)

FactoryTalk Historian To Historian Interface User Guide 15
time then exits. The times specified are relative to the machine where the interface runs. This
is important if the source Historian Server is in a different time zone than the machine where
the interface runs. Note that backfilling data may require additional server preparations on the
receiving node. For example, there may not be enough space in the target (non-primary)
archive for the recovery data, or non-primary archives may not have space allocated for
newly created tags. See the Historian Server System Management Guide for information on
backfilling data.
Exception Data Collection
Tags assigned to the first scan class receive exception data. An exception is a current value
update (snapshot value). On startup each source tag is registered with the update manager on
the source FactoryTalk Historian System. The update manager then collects snapshot updates
for each registered tag. This data is queued in memory. When the interface executes a scan it
requests and processes these events until the queue is empty. Data update latency is
minimized by configuring a high frequency scan rate for the exception scan class. Exception
and compression should be configured carefully for the interface to avoid data mismatches
between Historian Servers. See the section
Exception and Compression for additional
information.
Maximizing Data Throughput
Interface performance is maximized when a separate copy of the interface is used for archive
and exception data collection.
The number of events returned per update request to the source Historian Server is
configurable through the interface startup file. The default is for up to 10,000 exception
events returned per update request. This list of exceptions is given to the interface in time
sequence, oldest to newest. The interface processes the list one event at a time. These events
are temporarily queued by the interface. Only one update is queued per tag at any given time.
Whenever a second event for any one tag is processed within the exception update list, the
interface must first flush its queue, writing this data to the receiving server. This behavior is
required for exception data filtering. As a result, the interface will make many more calls
writing data to the receiving Historian Server than retrieving data from the source.
Note: Whenever possible, users should run the interface on the receiving Historian
Server to minimize the effect of network latency on data throughput.
Network latency will have a significant effect on data throughput. Interface testing over a
WAN connection with 200ms latency showed throughput was reduced by 2/3 when the
interface ran on the source versus the receiving Historian Server. When run on the receiving
Historian Server, the interface was able to sustain 1800 events/second. This was reduced to
600 events/second when the interface ran on the source Historian Server. Testing with a
latency of <1 ms, throughput was not affected. The interface maintained 11,000
events/second when run on the source and receiving Historian Server.