Review cost threshold for parallelism with MAXDOP, inspect NUMA layout and settings, then measure plans and waits after changes.
CXSYNC_PORT and CXSYNC_CONSUMER Waits in Parallel Queries
Investigate CXSYNC_PORT with session wait deltas, parallel plan exchanges and thread rows before testing a query-level MAXDOP change.
DOP Feedback: Letting SQL Server Lower Parallelism by Itself
Configure DOP feedback in SQL Server 2022, check Query Store evidence, and compare parallelism changes with your workload and MAXDOP limits.
Batch vs Stream Processing: When Data Cannot Wait
Batch vs Stream Processing is a choice about how long an answer can wait. This post compares an hourly batch job with a 5-minute stream window on the same six orders, covers the cost of real time and shows how to choose.
Bitmap Filters in Parallel Hash Joins: Reading Them in a Plan
Read bitmap filters in parallel hash join plans, compare fact scan output, and distinguish early row rejection from guaranteed savings.
MapReduce Explained: Map, Shuffle and Reduce Step by Step
MapReduce splits a big job into many small map tasks, moves the results by key in a shuffle, and adds them up in reduce tasks. This post walks one word count through every step, with a recreated diagram and the same idea in T-SQL.
How Apache Spark Works: Driver, Executors and Partitions
How Apache Spark works comes down to a driver that plans the job, executors that run it, and data cut into partitions. This post follows one sales total through two stages and a shuffle, and compares it with a parallel plan in SQL Server.