Incremental Parsing

Incremental parsing improves replication performance especially for replication environments with multiple Replication Servers in the replication route, and reduces memory consumption for HVAR, RTL, and DSI bulk-copy-in.

The single DSI Scheduler (DSI-S) thread parses data and then profiles transactions to determine whether to use the strategy of DSI bulk-copy-in, high volume adaptive replication (HVAR), or real-time loading (RTL) to apply transactions to the replicate database. DSI-S then conveys the parsed data and profiled transactions to one or more DSI Executor (DSI-E) threads that wait, in turn, for the single DSI-S thread to process data. To avoid a second parsing by the DSI-E thread to profile the data for the replicate database, Replication Server stores the entire parsed data in the SQT cache. However, this requires a large amount of memory, since Replication Server does not delete the data until more transactions are available to optimally compile transactions for HVAR or RTL, or to group transactions for DSI bulk-copy-in.

Incremental parsing allows the DSI-S thread to parse only table-level information, which is the first part of any DML command that the DSI-S thread processes. The DSI-S thread transfers the parsing tasks for the second part of the command to multiple DSI-E threads. With incremental parsing:
Set dsi_incremental_parsing to on to enable incremental parsing by the DSI-S thread when you enable high volume adaptive replication (HVAR), real-time loading (RTL), or DSI bulk-copy-in.
Note: dsi_incremental_parsing takes effect only if either dsi_compile_enable or dsi_bulk_copy is also set to on. Otherwise, Replication Server ignores dsi_incremental_parsing.

Default: on

The replicate Replication Servers must be version 15.7.1 SP100 or later to support incremental parsing.

Use dsi_incremental_parsing with: Connection-level settings override server-level settings.

See Real-Time Loading Solution in the Replication Server Heterogeneous Guide.

Related concepts
DSI Bulk Copy-in
High Volume Adaptive Replication to Adaptive Server