Processing of Timeline data

Timeline is aware of the concept of ownership. When a RealTime collect buffer is flushed to the Historical database (HDB), only rows that are owned by the local system are written to the database.

If a given system is interested in a point that is configured for collection, then real-time data is collected. This allows you to see "current" working values on a non-owning system even though the final data has not been written to the HDB.

When the mode switch results in the "ownership" of a dataset changing from one system to another, each system marks the affected collect rows for ‘flush’ operations.

On the Historical side, replication ensures that data is consistent across multiple systems. Replication services are aware of which systems are interested in a given dataset so that data can be correctly replicated when it is needed. The figure below illustrates the paths of replication required when several systems share datasets.

System 1 is the owner of datasets 1-6 and the other three systems are interested in the same data, but are not owners. For example, System 2 is interested in the points that belong to datasets 1 through 3, System 3 is interested in datasets 2, 4, and 5, and System 4 is interested in datasets 1 and 6. In this scenario, collect history rows for dataset 1 must be replicated to both System 2 and System 4.

The RDBMS is responsible for allocating the tag ID that uniquely defines the relationship between a RealTime database (RTDB) point and the rows of data in the Timeline tables. To ensure that tag ID values do not collide across systems, a fence table is used. This table, shown below, specifies the range of values that a system may allocate when an allocation operation is required.

xisTagRanges

Data conflicts can occur when two or more systems believe that they have active ownership of a specific point at given time. This may result in data being written to the same time in Timeline from two different sources. The primary cause of collisions of this nature are Wide Area Network (WAN) outages.

Caution: During a WAN outage, adding points to collect could cause synchronization problems. This is because systems on each side of the WAN may believe that they have ownership of the datasets involved.

Since WAN configurations are often non-redundant, a failure is likely to result in one or more systems being physically isolated from the others. You will have to determine which data to keep or to remove from a system.

Also, the RealTime database (RTDB) Collect processes that are responsible in allocating tag IDs only do so to a given point if the system has ownership of that point. This prevents two systems from allocating tag IDs concurrently for a point, since only one system at a time can have ownership.