Dapper · Dapper.SimpleCRUD · EF Core

The same table and the same operations on every database: first the libraries on their own, then through an instant OData/REST API built on each. Lower is better. Generated 2026-09-30 21:50 UTC.

OS macOS 26.6.2arch Arm64cores 16.NET .NET 10.0.1Dapper 2.0.78Dapper.SimpleCRUD 2.3.0EF Core 10.0.0ez-odata-api 1.0.7EzOdata.SimpleCrud 2.0.1drivers Npgsql 10.0.0, MySqlConnector 2.3.7, SqlClient 6.1.1, Microsoft.Data.Sqlite 10.0.0databases SQLite, PostgreSQL, MySQL, SQL Server

Dapper.SimpleCRUD against hand-written Dapper (geometric mean across databases): get by id 1.24×, filtered page (20 rows) 1.10×, insert one 1.04×, update one 1.10×, insert + delete 1.05×, insert 100 (one transaction) 1.38×.
EF Core against hand-written Dapper (geometric mean across databases): get by id 1.60×, filtered page (20 rows) 1.41×, insert one 1.21×, update one 1.20×, insert + delete 1.15×, insert 100 (one transaction) 0.30×.
The EzOdata facade (SimpleCRUD in an isolated copy per dialect) against SimpleCRUD's static API on SQLite: get by id +1%, filtered page (20 rows) +1%, insert one +2%, update one -6%, insert + delete -2%, insert 100 (one transaction) +0%; 6 of 6 are inside the confidence intervals.
EF Core change tracking on reads, relative to Dapper: get by id 1.60× tracked vs 1.61× AsNoTracking, filtered page (20 rows) 1.41× tracked vs 1.35× AsNoTracking.
Allocations per operation, SimpleCRUD / EF Core relative to Dapper: get by id 3.4× / 2.8×, filtered page (20 rows) 2.2× / 3.4×, insert one 1.7× / 3.1×, update one 2.7× / 3.4×, insert + delete 1.8× / 3.6×, insert 100 (one transaction) 3.2× / 3.6×.
Reads through the HTTP API take the same path on every engine (ez-odata's compiled SQL); median difference from stock ez-odata, SimpleCRUD / EF Core engine: get-by-key +0.01 ms / +0.01 ms, list +0.01 ms / 0.00 ms, expand 0.00 ms / 0.00 ms.
Writes through the HTTP API (typed entity, hooks, one transaction), median difference from stock ez-odata, SimpleCRUD / EF Core engine: post +0.30 ms / +1.04 ms, patch +0.63 ms / +1.44 ms, put +0.58 ms / +0.36 ms.
Tests: 71 passed, 0 failed, 0 skipped.

The libraries on their own

No HTTP and no ez-odata: each library does the operation the way its documentation recommends, opening a pooled connection per operation as a web request would. Bars are the mean, whiskers the 99.9% confidence interval; on the right, the mean (with the ratio to Dapper) and the bytes allocated. Hover a bar for the median and sample count.

Dapper, hand-written SQLDapper.SimpleCRUD (EzOdata facade)Dapper.SimpleCRUD (static API)EF CoreEF Core, AsNoTracking

Get by id

SQLite

Dapper5.1 µs
3.1 KB

SimpleCRUD12.0 µs 2.35×
12.9 KB

SimpleCRUD static11.9 µs 2.32×
12.9 KB

EF Core18.3 µs 3.59×
12.1 KB

EF no-tracking18.6 µs 3.64×
12.6 KB

PostgreSQL

Dapper324 µs
3.5 KB

SimpleCRUD342 µs 1.06×
13.5 KB

EF Core358 µs 1.11×
9.3 KB

EF no-tracking349 µs 1.08×
9.8 KB

MySQL

Dapper606 µs
8.0 KB

SimpleCRUD584 µs 0.96×
17.5 KB

SQL Server

Dapper703 µs
8.0 KB

SimpleCRUD704 µs 1.00×
17.9 KB

EF Core724 µs 1.03×
14.1 KB

EF no-tracking741 µs 1.05×
14.6 KB

Filtered page (20 rows)

SQLite

Dapper18.1 µs
8.1 KB

SimpleCRUD26.0 µs 1.44×
19.2 KB

SimpleCRUD static25.7 µs 1.43×
19.2 KB

EF Core41.8 µs 2.32×
32.0 KB

EF no-tracking37.0 µs 2.05×
23.7 KB

PostgreSQL

Dapper418 µs
8.5 KB

SimpleCRUD413 µs 0.99×
19.7 KB

EF Core478 µs 1.14×
29.3 KB

EF no-tracking455 µs 1.09×
21.0 KB

MySQL

Dapper676 µs
13.0 KB

SimpleCRUD671 µs 0.99×
23.4 KB

SQL Server

Dapper750 µs
13.2 KB

SimpleCRUD788 µs 1.05×
25.7 KB

EF Core793 µs 1.06×
35.1 KB

EF no-tracking832 µs 1.11×
26.8 KB

Insert one

SQLite

Dapper44.5 µs
4.9 KB

SimpleCRUD50.5 µs 1.13×
8.4 KB

SimpleCRUD static49.5 µs 1.11×
8.4 KB

EF Core62.5 µs 1.41×
16.1 KB

PostgreSQL

Dapper335 µs
4.0 KB

SimpleCRUD344 µs 1.03×
7.2 KB

EF Core414 µs 1.23×
14.1 KB

MySQL

Dapper1.12 ms
6.6 KB

SimpleCRUD1.08 ms 0.97×
10.2 KB

SQL Server

Dapper1.12 ms
8.4 KB

SimpleCRUD1.16 ms 1.03×
11.9 KB

EF Core1.14 ms 1.02×
19.6 KB

Update one

SQLite

Dapper37.1 µs
3.7 KB

SimpleCRUD44.8 µs 1.21×
11.6 KB

SimpleCRUD static47.5 µs 1.28×
11.6 KB

EF Core55.0 µs 1.48×
16.0 KB

PostgreSQL

Dapper352 µs
4.4 KB

SimpleCRUD386 µs 1.10×
12.4 KB

EF Core385 µs 1.09×
14.0 KB

MySQL

Dapper1.07 ms
6.4 KB

SimpleCRUD1.11 ms 1.04×
14.1 KB

SQL Server

Dapper1.08 ms
7.3 KB

SimpleCRUD1.16 ms 1.08×
15.3 KB

EF Core1.14 ms 1.06×
19.3 KB

Insert + delete

SQLite

Dapper82.4 µs
6.7 KB

SimpleCRUD86.9 µs 1.05×
12.9 KB

SimpleCRUD static89.0 µs 1.08×
12.9 KB

EF Core109 µs 1.32×
29.4 KB

PostgreSQL

Dapper669 µs
6.3 KB

SimpleCRUD713 µs 1.07×
12.2 KB

EF Core721 µs 1.08×
24.6 KB

MySQL

Dapper2.06 ms
11.7 KB

SimpleCRUD2.07 ms 1.00×
18.0 KB

SQL Server

Dapper2.13 ms
13.5 KB

SimpleCRUD2.32 ms 1.09×
19.5 KB

EF Core2.28 ms 1.07×
35.0 KB

Insert 100 (one transaction)

SQLite

Dapper318 µs
185.7 KB

SimpleCRUD979 µs 3.08×
750.7 KB

SimpleCRUD static975 µs 3.06×
744.4 KB

EF Core1.10 ms 3.46×
923.1 KB

PostgreSQL

Dapper31.40 ms
197.4 KB

SimpleCRUD33.44 ms 1.06×
553.0 KB

EF Core2.63 ms 0.08×
768.7 KB

MySQL

Dapper23.70 ms
262.7 KB

SimpleCRUD24.91 ms 1.05×
762.3 KB

SQL Server

Dapper64.20 ms
404.0 KB

SimpleCRUD67.99 ms 1.06×
1.04 MB

EF Core6.02 ms 0.09×
780.3 KB

Through the HTTP API

An in-memory TestServer per engine: stock ez-odata (untyped), the SimpleCRUD engine and the EF Core engine (typed entities, hooks, a transaction per write). For each database and scenario the engines take turns request by request after a shared warm-up. Median / p95 in ms; bytes allocated per request across the whole process.

stock ez-odataSimpleCRUD engineEF Core engine

SQLite

scenario	stock	simplecrud	efcore
post	0.53 / 0.74 56.5 KB	0.56 / 0.76 64.8 KB	0.76 / 1.06 128.8 KB
get-by-key	0.14 / 0.21 53.1 KB	0.14 / 0.18 54.9 KB	0.15 / 0.19 55.2 KB
list	0.25 / 0.33 98.2 KB	0.26 / 0.34 99.0 KB	0.25 / 0.33 99.1 KB
expand	0.22 / 0.42 80.7 KB	0.23 / 0.44 83.0 KB	0.22 / 0.36 82.6 KB
patch	0.49 / 0.94 55.8 KB	0.55 / 0.97 79.9 KB	0.80 / 1.38 141.0 KB
put	0.19 / 0.26 58.3 KB	0.22 / 0.30 80.1 KB	0.35 / 0.49 132.7 KB
delete-post	0.87 / 1.40 89.3 KB	0.93 / 1.42 117.1 KB	1.31 / 2.36 243.4 KB
concurrent-16	1659.49 / 2114.41 921.7 KB	1516.28 / 2113.67 1.02 MB	1667.12 / 2115.38 2.00 MB
hook-audit	—	0.30 / 0.45 68.0 KB	0.38 / 0.56 137.5 KB
direct-insert	—	0.31 / 0.39 7.8 KB	0.35 / 0.47 70.5 KB

PostgreSQL

scenario	stock	simplecrud	efcore
post	1.00 / 1.26 60.8 KB	1.41 / 1.86 69.5 KB	2.22 / 2.81 131.1 KB
get-by-key	0.56 / 0.70 59.3 KB	0.57 / 0.75 61.5 KB	0.57 / 0.72 61.0 KB
list	0.72 / 0.83 103.8 KB	0.73 / 0.84 105.1 KB	0.72 / 0.85 105.6 KB
expand	1.09 / 1.74 91.3 KB	1.09 / 1.82 93.3 KB	1.08 / 1.66 92.7 KB
patch	0.89 / 1.36 59.9 KB	1.65 / 2.63 84.2 KB	2.58 / 3.86 142.4 KB
put	0.79 / 0.97 62.1 KB	1.49 / 1.88 84.9 KB	1.31 / 1.63 129.9 KB
delete-post	1.55 / 1.89 98.5 KB	2.28 / 2.94 125.5 KB	3.72 / 4.42 242.3 KB
concurrent-16	1.91 / 3.65 1.11 MB	2.60 / 4.64 1.19 MB	4.62 / 7.60 2.11 MB
hook-audit	—	1.95 / 3.25 71.1 KB	3.62 / 6.23 142.1 KB
direct-insert	—	0.39 / 0.53 6.2 KB	0.44 / 0.62 62.4 KB

MySQL

scenario	stock	simplecrud	efcore
post	1.98 / 2.43 69.6 KB	2.04 / 2.48 73.2 KB	—
get-by-key	0.82 / 0.97 61.0 KB	0.83 / 0.97 62.5 KB	—
list	1.03 / 1.43 105.9 KB	1.04 / 1.46 106.3 KB	—
expand	1.58 / 1.87 93.2 KB	1.59 / 1.91 96.2 KB	—
patch	1.87 / 3.19 67.8 KB	2.24 / 3.88 90.1 KB	—
put	1.55 / 2.24 70.0 KB	1.90 / 2.96 90.5 KB	—
delete-post	3.52 / 7.14 106.9 KB	3.97 / 9.21 133.5 KB	—
concurrent-16	4.62 / 6.80 1.13 MB	4.65 / 7.15 1.21 MB	—
hook-audit	—	2.63 / 3.41 76.3 KB	—
direct-insert	—	1.11 / 1.43 9.1 KB	—

SQL Server

scenario	stock	simplecrud	efcore
post	2.61 / 4.99 64.4 KB	3.30 / 6.22 75.5 KB	4.27 / 7.69 173.2 KB
get-by-key	0.82 / 1.34 59.8 KB	0.83 / 1.45 61.5 KB	0.84 / 1.40 60.9 KB
list	0.97 / 1.27 104.5 KB	0.99 / 1.17 104.2 KB	0.98 / 1.28 104.9 KB
expand	1.52 / 1.77 92.1 KB	1.51 / 1.79 94.1 KB	1.52 / 1.84 93.9 KB
patch	2.47 / 2.89 63.4 KB	3.81 / 4.41 94.8 KB	4.77 / 5.50 187.6 KB
put	2.50 / 3.59 65.4 KB	3.75 / 5.53 95.4 KB	2.90 / 3.90 170.5 KB
delete-post	5.10 / 6.49 101.8 KB	6.60 / 8.94 137.9 KB	8.50 / 10.81 326.2 KB
concurrent-16	8.97 / 16.03 1.07 MB	11.62 / 20.28 1.23 MB	14.77 / 26.40 2.72 MB
hook-audit	—	4.23 / 5.39 82.7 KB	5.85 / 7.67 189.9 KB
direct-insert	—	1.22 / 1.45 10.9 KB	1.32 / 1.58 101.6 KB

Tests

group	passed	time
EF Core engine suites	18	2.2 s
facade, engines and API	35	3.7 s
SimpleCRUD engine suites	18	1.9 s

Every test

EzOdata.Entities.AspNetCore.Tests.EfCore_CompositeKeyTests.Composite_keys_work_end_to_end	Passed	215 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_MultiDialectTests.Every_dialect_behind_one_api_in_one_process	Passed	1539 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_PropertyNameTests.Entity_property_names_become_the_api_contract	Passed	209 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_PropertyNameTests.Row_filters_and_expand_follow_the_renamed_columns	Passed	2 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Columns_the_entity_does_not_write_are_read_only_in_the_api	Passed	5 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Database_constraint_violations_map_to_the_engines_error_codes	Passed	4 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Delete_can_be_overridden_as_a_soft_delete	Passed	74 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Expand_and_after_read_hooks_work_on_extended_tables	Passed	2 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Handlers_can_forbid_an_update_based_on_the_original_row	Passed	5 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Hook_rejection_is_a_400_and_nothing_is_written	Passed	2 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Insert_outside_the_callers_row_filter_is_403_and_rolled_back	Passed	2 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Insert_runs_through_the_engine_and_the_handler	Passed	12 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.OnCommitted_runs_once_after_commit_and_never_after_rollback	Passed	86 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Patch_changes_only_what_was_sent	Passed	3 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Reads_apply_role_row_filters_and_handler_filters_together	Passed	2 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Row_filters_protect_updates_and_deletes	Passed	11 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Side_writes_in_hooks_roll_back_with_the_api_write	Passed	5 ms
EzOdata.Entities.AspNetCore.Tests.EfCore_SqliteEndToEndTests.Tables_without_an_entity_keep_the_stock_instant_api	Passed	4 ms
EzOdata.Entities.AspNetCore.Tests.EfCoreGuardTests.Inheritance_mapping_is_rejected	Passed	16 ms
EzOdata.Entities.AspNetCore.Tests.EfCoreGuardTests.Owned_types_are_rejected	Passed	75 ms
EzOdata.Entities.AspNetCore.Tests.EfCoreGuardTests.Retrying_execution_strategy_is_rejected	Passed	559 ms
EzOdata.Entities.AspNetCore.Tests.EfCoreGuardTests.Unsupported_key_types_are_rejected	Passed	6 ms
EzOdata.Entities.AspNetCore.Tests.EfCoreGuardTests.Value_converters_produce_a_startup_warning	Passed	25 ms
EzOdata.Entities.AspNetCore.Tests.EscapeHatchTests.EfCore_escape_hatch_shares_the_write_transaction	Passed	201 ms
EzOdata.Entities.AspNetCore.Tests.EscapeHatchTests.SimpleCRUD_escape_hatch_shares_the_write_transaction	Passed	483 ms
EzOdata.Entities.AspNetCore.Tests.OpenApiTests.Openapi_server_urls_point_at_the_service	Passed	493 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_CompositeKeyTests.Composite_keys_work_end_to_end	Passed	165 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_MultiDialectTests.Every_dialect_behind_one_api_in_one_process	Passed	1295 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_PropertyNameTests.Entity_property_names_become_the_api_contract	Passed	167 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_PropertyNameTests.Row_filters_and_expand_follow_the_renamed_columns	Passed	37 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Columns_the_entity_does_not_write_are_read_only_in_the_api	Passed	9 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Database_constraint_violations_map_to_the_engines_error_codes	Passed	3 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Delete_can_be_overridden_as_a_soft_delete	Passed	37 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Expand_and_after_read_hooks_work_on_extended_tables	Passed	3 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Handlers_can_forbid_an_update_based_on_the_original_row	Passed	6 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Hook_rejection_is_a_400_and_nothing_is_written	Passed	2 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Insert_outside_the_callers_row_filter_is_403_and_rolled_back	Passed	4 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Insert_runs_through_the_engine_and_the_handler	Passed	4 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.OnCommitted_runs_once_after_commit_and_never_after_rollback	Passed	8 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Patch_changes_only_what_was_sent	Passed	7 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Reads_apply_role_row_filters_and_handler_filters_together	Passed	2 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Row_filters_protect_updates_and_deletes	Passed	3 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Side_writes_in_hooks_roll_back_with_the_api_write	Passed	11 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrud_SqliteEndToEndTests.Tables_without_an_entity_keep_the_stock_instant_api	Passed	140 ms
EzOdata.Entities.AspNetCore.Tests.SimpleCrudNamingTests.UseNaming_maps_snake_case_tables	Passed	475 ms
EzOdata.Entities.AspNetCore.Tests.StartupValidationTests.An_entity_that_does_not_match_the_database_stops_startup	Passed	358 ms
EzOdata.Entities.AspNetCore.Tests.StartupValidationTests.Composite_key_parts_must_be_Required_for_SimpleCRUD	Passed	8 ms
EzOdata.Entities.AspNetCore.Tests.StartupValidationTests.ExtendEzOData_requires_AddEzOData_first	Passed	1 ms
EzOdata.Entities.AspNetCore.Tests.StartupValidationTests.Missing_engine_stops_startup	Passed	5 ms
EzOdata.SimpleCrud.Tests.EngineTests.An_incompatible_SimpleCRUD_is_rejected_with_a_clear_message	Passed	0 ms
EzOdata.SimpleCrud.Tests.EngineTests.Delegates_are_compiled_once_per_operation_and_type	Passed	5 ms
EzOdata.SimpleCrud.Tests.EngineTests.Describe_reads_the_mapping_from_SimpleCRUD_itself	Passed	1 ms
EzOdata.SimpleCrud.Tests.EngineTests.Each_engine_emits_its_own_dialect_in_one_process(dialect: MySQL, table: "`customers`", id: "`Id`", paging: "LIMIT 0,10")	Passed	14 ms
EzOdata.SimpleCrud.Tests.EngineTests.Each_engine_emits_its_own_dialect_in_one_process(dialect: PostgreSQL, table: "\"customers\"", id: "\"Id\"", paging: "LIMIT 10 OFFSET ((1-1) * 10)")	Passed	69 ms
EzOdata.SimpleCrud.Tests.EngineTests.Each_engine_emits_its_own_dialect_in_one_process(dialect: SQLite, table: "\"customers\"", id: "\"Id\"", paging: "LIMIT 10 OFFSET ((1-1) * 10)")	Passed	9 ms
EzOdata.SimpleCrud.Tests.EngineTests.Each_engine_emits_its_own_dialect_in_one_process(dialect: SQLServer, table: "[customers]", id: "[Id]", paging: "ROW_NUMBER() OVER(ORDER BY full_name)")	Passed	14 ms
EzOdata.SimpleCrud.Tests.EngineTests.Engines_are_lazy_singletons_per_dialect	Passed	0 ms
EzOdata.SimpleCrud.Tests.EngineTests.Exceptions_surface_unwrapped	Passed	4 ms
EzOdata.SimpleCrud.Tests.EngineTests.First_use_under_contention_creates_exactly_one_engine	Passed	4 ms
EzOdata.SimpleCrud.Tests.EngineTests.Isolated_engines_never_touch_the_process_wide_SimpleCRUD	Passed	7 ms
EzOdata.SimpleCrud.Tests.EngineTests.Keyed_DI_registrations_resolve_independent_dialects	Passed	12 ms
EzOdata.SimpleCrud.Tests.EngineTests.Naming_conventions_are_per_engine	Passed	17 ms
EzOdata.SimpleCrud.Tests.EngineTests.Overhead_versus_calling_SimpleCRUD_directly	Passed	348 ms
EzOdata.SimpleCrud.Tests.EngineTests.Sessions_commit_and_roll_back	Passed	11 ms
EzOdata.SimpleCrud.Tests.EngineTests.Several_custom_naming_engines_coexist	Passed	14 ms
EzOdata.SimpleCrud.Tests.EngineTests.Shared_engine_never_changes_global_state_implicitly	Passed	1 ms
EzOdata.SimpleCrud.Tests.EngineTests.Why_isolation_is_needed_switching_SimpleCRUDs_dialect_leaves_stale_sql	Passed	10 ms
EzOdata.SimpleCrud.Tests.MultiDatabaseTests.Every_database_at_once_in_one_process	Passed	71 ms
EzOdata.SimpleCrud.Tests.MultiDatabaseTests.Full_crud_through_the_facade(kind: "mysql")	Passed	117 ms
EzOdata.SimpleCrud.Tests.MultiDatabaseTests.Full_crud_through_the_facade(kind: "postgresql")	Passed	126 ms
EzOdata.SimpleCrud.Tests.MultiDatabaseTests.Full_crud_through_the_facade(kind: "sqlserver")	Passed	107 ms

How this was measured

Correctness first. Before any timing, every library reads the same 20-row page, then inserts, reads back, updates and deletes a row, on every database; the run stops on any difference.
One database at a time. Each database runs in its own process with only its own container started (PostgreSQL 16, MySQL 8.4, SQL Server 2022, or Azure SQL Edge on ARM), so they do not compete for the Docker VM. SQLite is a local file in WAL mode.
Library benchmarks use BenchmarkDotNet in process: 5 warm-up and 20 measured iterations of about 250 ms each, with the memory diagnoser. Each (library, database) case starts from a freshly created table with 5,000 rows.
Same work for everyone. Dapper runs hand-written SQL and reads the new id back (RETURNING, OUTPUT or LAST_INSERT_ID). SimpleCRUD uses Get, GetListPaged, Insert, Update and Delete. EF Core uses a pooled DbContext factory with one context per operation, as ASP.NET Core does. Insert 100 is one transaction: SimpleCRUD inserts row by row and returns every id, EF Core batches, and Dapper's list form runs row by row without returning ids.
SimpleCRUD twice. The EzOdata facade runs SimpleCRUD in an isolated copy per dialect; SimpleCRUD's own static API is measured next to it on SQLite (its dialect is process-wide, so one database only) to show what the facade costs.
EF Core × MySQL is not measured: there is no MySqlConnector-based EF Core 10 provider yet.
SQLite has one writer. Its concurrent-16 row measures SQLite's lock retries (about 1.6 s per batch of 16 inserts), not the engines.
HTTP numbers come from an in-memory TestServer (no sockets), so they show the API layer plus the database, not the network. Engines are interleaved request by request so that none benefits from running later; allocations are whole-process bytes per request.
Microbenchmarks on one machine: treat differences inside the confidence interval as equal, and rerun with ./try.sh --benchmark on your own hardware.

Raw results: raw/ (BenchmarkDotNet tables and HTTP timings per database).