Wait stats over time means measuring the waits of one window, such as your busy hour, instead of reading totals since the last restart. On Night 5 at the Clipboard Diner, Casey photographs the clipboard at 6:00 and at 7:00 and subtracts. Then one script does the same thing with two snapshots of sys.dm_os_wait_stats.
sys.dm_os_wait_stats: Reading the Wait Stats List
sys.dm_os_wait_stats is the running list of every wait SQL Server has recorded since its last restart. On Night 3 at the Clipboard Diner, Casey reads the clipboard at 2 AM and crosses out the sleepers. Then one query filters out the harmless waits and shows the share, the average and the signal part of every wait that matters.
Wait Stats Explained: SQL Server as an All-Night Diner
Wait stats explained with one picture: SQL Server is an all-night diner, and every query is an order ticket that waits for something. Meet the Clipboard Diner, its owner Casey, the cooks and booth 7. Then count the burners and cooks on your own server with one query.
Index Write Cost: What Each New Index Adds to an INSERT
Index write cost is the part of the create-an-index advice that the cartoon leaves out. This test inserts the same 20,000 rows with no extra index, one and three, and measures the log written each time. It also shows what the index saved on a lookup, so you can weigh both sides.
When Auto Update Statistics Is Not Enough
Auto update statistics can miss important data patterns; inspect thresholds, ascending keys, sampling, and targeted update schedules.
LIKE Cardinality: String Statistics and Row Estimates
LIKE row estimates depend on the pattern, string statistics and compilation context. Compare estimated and actual matching rows in a controlled distribution, while keeping cardinality estimates separate from the chosen access path.
Setting MAXDOP and Cost Threshold Sensibly
MAXDOP and cost threshold work together to shape parallel plans; inspect current values, test a change, and watch real workload effects.