SQL Server Consultants in Tampa, FL

What the review found. The reporting query filtered a large table with no covering index behind it, so every call scanned the table end to end, about 141,545 reads apiece.

What we changed (identifiers generalized for privacy):

-- Reporting query scanned the whole table: 141,545 logical reads, ~189 ms per run.
CREATE NONCLUSTERED INDEX IX_sample_results_covering
    ON dbo.[sample_results] (/* filter cols */) INCLUDE (/* output cols */);
-- rollback: DROP INDEX IX_sample_results_covering provided

The covering index flipped the scan into a seek: reads fell from 141,545 to 9, and the run dropped from about 189 ms to roughly 1 ms.

What the audit turned up. One application database was consuming about 91.81% of the instance's CPU and 92.17% of its storage I/O, the pressure was overwhelmingly CPU-bound with CXPACKET at 71.8% of waits, and the data drive was writing at 1,247 ms against a 20 ms healthy mark.

What the report flagged first (findings summarized, identifiers withheld):

-- Red9 145-point health check: resource & storage exposure
-- P1  cpu       : one application DB at ~91.81% of instance CPU
-- P1  io        : same DB at ~92.17% of storage I/O
-- P1  waits     : CXPACKET 71.8% of all waits (parallelism)
-- P2  storage   : data-drive write latency ~1,247 ms vs ~20 ms target

Each item came with its remediation step and rollback beside it, so the company could take the CPU and parallelism findings first and schedule the storage work against its own calendar.

What the review turned up. A production instance had seven databases with no backup inside seven days, 132 deadlocks across a four-day window, 1,043 transaction-log VLFs, tempdb near 60% of usage, and a memory dump on record.

What the report flagged first (findings summarized, identifiers withheld):

-- Red9 145-point health check: recoverability & stability exposure
-- P1  backups   : 7 databases with no backup in the last 7 days
-- P1  deadlocks : 132 in a 4-day window (~33/day)
-- P2  vlfs      : 1,043 transaction-log VLFs (slow recovery/startup)
-- P2  tempdb    : ~60% usage, plus a recorded memory dump

Each finding shipped with its remediation and rollback beside it, so the company could put the seven databases on a backup schedule first and work the deadlocks, VLFs, and tempdb pressure against its own calendar.