Autonomous Performance Engineering · Windows Server 2022 · SQL Server 2022
| ID & Priority | Recommendation Title | Category | Estimated Impact | Risk Level | Lifecycle Status | Action |
|---|
SQL Server CPU pressure spiked +31% over the baseline. 78% of the CPU saturation is correlated with query fingerprint USP_GetCustomerPortfolioSummary, which regressed from an Index Seek to a Clustered Table Scan on August 22. Deploying covering index IX_Transactions_CustID_Date_Amount will reduce duration from 18.2s to 4.0s (77.8% gain).
Self-service workflow to register, connect, baseline, and activate autonomous monitoring for any SQL Server 2022 / Windows environment.
Organization, Tier, and Technical Contacts.
Production, UAT, Staging, or Restricted Nexus.
PerfMon WMI / WinRM agentless connection.
Read-only monitoring credentials (VIEW SERVER STATE).
Latency check, port 1433 & DMV availability.
Non-destructive baseline performance capture.
| Counter | Current Value | Status |
|---|---|---|
| Processor: % Processor Time | 92.4% | Saturated |
| Memory: Page Life Expectancy | 82 sec | Low PLE |
| LogicalDisk(P102): Avg Disk sec/Read | 48.2 ms | Slow I/O |
| LogicalDisk(P102): Avg Disk sec/Write | 12.4 ms | Optimal |
| LogicalDisk: Current Queue Length | 4.8 | Queued |
| SQLServer:SQL Statistics: Batch Req/sec | 3,420 | Normal |
| Wait Type | % Total Waits | Analysis & Impact |
|---|
Performs controlled non-destructive baseline assessments to establish performance truth.
| Query ID | Query Name / Stored Procedure | Exec/Hr | Duration | Logical Reads | Plan Regressed? | Severity | Action |
|---|
Comparison of Original regressed query vs Proposed sargable rewrite with DDL index script.
Paste any SQL query to estimate CPU time, logical reads, anti-patterns, and production risk before pushing to production.
Enforces safety rules to prevent runaway queries from compromising production availability.
| Policy ID | Policy Name | Monitored Metric | Threshold | Enforcement Action | Status |
|---|
Multi-signal correlation mapping the complete causal degradation chain.
Sustained CPU is only 34%. 92% peak is driven by table scan query plan regression. Upsizing vCPUs would add $14k/yr licensing without solving the scan.
Host OS has 5.8 GB free. Page Life Expectancy drop to 82s is due to cache eviction from the 1.42M read query, not memory sizing.
Throughput capacity is adequate, but shared TempDB creates I/O queue lockup. Separate TempDB onto dedicated fast volume VMDK P103.
| Category | Standard Rule Description | Current Setting | Recommended Target | Compliance |
|---|
Interactive training courses generated from anonymized production performance regressions.
| Specialized Agent Name | Core Domain & Responsibilities | Status | Latency | Processed Events |
|---|