Expand query execution coverage in the performance test suite

[benrr101]

The existing benchmarks cover a limited set of command and reader scenarios, and some include fixture preparation or measure iterations too short for stable results. This change expands coverage within the existing runners and focuses measurements on command execution and result consumption.

Changes

• Cover synchronous and asynchronous command APIs across SQL text, parameterized queries, and stored-procedure RPCs.
• Add typed getters and GetValues alongside reader drain baselines, including mixed 16- and 64-column rows with NULL values.
• Cover binary and Unicode large values through materialization, chunked getters, and streaming, including 128 MiB plaintext payloads and both Default and SequentialAccess.
• Exercise MARS enabled and disabled, while retaining supported Always Encrypted scenarios.
• Move reusable commands, buffers, and fixture preparation outside measurement; validate fixtures and consumed results outside timing.
• Batch 512 sequential executions in fast command benchmarks, with an explanatory comment and OperationsPerInvoke normalization so results remain per command.

Infrastructure changes retain existing runner names, discovery, and configuration entry points. BenchmarkDotNet filtering and listing now work across enabled units, and invalid configurations or failed cases produce a failing exit code.

Validation

• Completed all 1,429 cases across 18 enabled units against a local SQL Server. The existing pool-contention runner required a retry with a longer case timeout.
• Reran all 48 command cases after batching: the shortest measured iteration was 194 ms, with no iteration-timing warnings.
• Command cases exceeding 20% variation dropped from 3 to 0; median variation dropped from 6.4% to 2.4%.
• Release build and all seven runner checks passed.

Measurement boundaries have changed, so driver comparisons should use the same updated benchmark sources rather than historical results from the previous implementations.

🤖

Implemented and validated entirely using 🤖

[Copilot]

Copilot review overview

🟡 Changes recommended

Timeout validation conflicts with existing semantics, and the documented regression-check script is missing.

Review effort: Balanced
Findings: 1 · 1

Open (2)

• Negative timeout values are rejected despite documented fallback · New
• Documented PowerShell regression script is missing · New

What changed in this PR

Expands performance benchmarks for command execution, typed readers, large payloads, MARS, and Always Encrypted scenarios.

Changes:

• Adds broader synchronous/asynchronous command and reader benchmarks.
• Moves fixture setup and allocations outside measured regions.
• Adds filtering, validation, and benchmark failure reporting.

Validation was source inspection only; tests were not run.

File	Description
runnerconfig.jsonc	Adds 128 MiB payload and longer timeouts.
README.md	Documents coverage and execution.
Program.cs	Adds filtering and failure handling.
DBFramework/​Table.cs	Adds setup timeouts and disposal.
DBFramework/​MixedRowFixture.cs	Creates mixed-row fixtures.
DBFramework/​DbUtils.cs	Applies command timeouts.
Config/​Loader.cs	Disables nullable analysis.
Config/​Config.cs	Introduces specialized job types.
Config/​CommandRunnerJob.cs	Adds defaults and validation.
SqlCommandRunner.cs	Expands command API benchmarks.
LargeDataReadRunner/​Plaintext.cs	Adds binary/text accessor coverage.
LargeDataReadRunner/​LargeDataReadRunnerBase.cs	Centralizes large-value setup and validation.
LargeDataReadRunner/​AlwaysEncrypted.cs	Adapts encrypted large-value benchmarks.
DataTypeReaderRunner/​ReaderCase.cs	Models individual and mixed reader cases.
DataTypeReaderRunner/​Plaintext.cs	Adds mixed rows and sequential access.
DataTypeReaderRunner/​DataTypeReaderRunnerBase.cs	Adds typed and GetValues benchmarks.
DataTypeReaderRunner/​AlwaysEncrypted.cs	Adapts encrypted reader benchmarks.
CommandRunnerBase.cs	Centralizes connection and command lifecycle.

---

💡 Configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🟡 Changes recommended

Mixed-row generation is nondeterministic, and the documented regression-check script is missing.

Review effort: Balanced
Findings: 1 · 1

Open (2)

• Make TOP selection deterministic in number generation · New
• Documented PowerShell regression script is missing

Resolved since last review (1)

• Negative timeout values are rejected despite documented fallback

[priyankatiwari08]

Partial review - diff is large, so I focused on SqlCommandRunner, CommandRunnerBase, DataTypeReaderRunnerBase, LargeDataReadRunnerBase, MixedRowFixture and Program.cs.

• Async readers/streams lost await using in six places - see inline.
• Program.cs: the first failing unit sets ExitCode = 1 and returns, so later enabled units are silently skipped; also BenchmarkRunner.Run(type, config, args) lets command-line args override the unit's job config that was just validated.

[priyankatiwari08]

This was await using before - plain using forces synchronous disposal of an async reader, which can block draining remaining results and skews what the async benchmark measures. Same downgrade in DataTypeReaderRunnerBase (3x), SqlCommandRunner, and the GetStream path.

[benrr101]

Sure, I'll just add those wherever valid.

[Copilot]

Copilot review overview

🟡 Changes recommended

The harness incorrectly fails for --info, and the README references a regression script that is not checked in.

Review effort: Balanced
Findings: 1 · 1

Open (2)

• Handle --info as a non-execution mode · New
• Documented PowerShell regression script is missing

Resolved since last review (1)

• Make TOP selection deterministic in number generation

[Copilot]

🔵 Needs a closer look

The extensive SQL Server, MARS, large-payload, and Always Encrypted matrix requires human review of independently executed benchmark results.

0 open findings

2 resolved since last review

• Handle --info as a non-execution mode
• Documented PowerShell regression script is missing

🧠 Review effort: Balanced