aljaz/turso
1
0
Fork 0
mirror of https://github.com/aljazceru/turso.git synced 2025-12-25 12:04:21 +01:00
Code Issues Packages Projects Releases Wiki Activity
10,048 Commits 1 Branch 0 Tags
14e104f8308ae87f0ba5600709c930e032d01027
Commit Graph

88 Commits

Author SHA1 Message Date
Pekka Enberg
d5c49c17c7 perf/throughput: Delete database before benchmark run 
Let's reduce variance between benchmark runs by starting with an empty
database.
 2025-10-07 09:57:08 +03:00
Avinash Sajjanshetty
36d4b09830 Use the updated bindings for encryption throughput tests 2025-10-02 19:04:15 +05:30
Pekka Enberg
981a762fd7 Merge 'Improve throughput benchmarks' from Pekka Enberg 
Closes #3493
 2025-10-01 15:24:03 +03:00
Pekka Enberg
51f4f1fb8b perf/throughput: Add plotting scripts 
This adds few helper scripts to plot throughput results.
 2025-10-01 14:08:26 +03:00
Pekka Enberg
3fcb0581ec perf/throughput: Fix thread pool size in Turso benchmark 
Replace #[tokio::main] with explicit Runtime builder to set the number
of tokio worker threads to match the benchmark thread count. This
ensures proper thread control and avoids interference from default
tokio thread pool sizing.
 2025-10-01 14:08:26 +03:00
Pekka Enberg
63895dfecd perf/throughput: Simplify benchmark output to CSV format 
Remove verbose output from rusqlite benchmark and output only CSV
format: system,threads,batch_size,compute,throughput

This makes it easier to parse and plot benchmark results.
 2025-10-01 11:06:27 +03:00
Pekka Enberg
eeb14b25c6 perf/throughput: Replace think time with CPU-bound compute time 
Replace the sleep-based --think parameter with a --compute parameter
that uses a busy loop to simulate realistic CPU or GPU bound business
logic (e.g., parsing, data aggregation, or ML inference). The compute
time is now specified in microseconds instead of milliseconds for
finer granularity.
 2025-10-01 11:06:27 +03:00
Jussi Saurio
d2863dd62f Merge 'Measure read/write latencies in encryption benchmarks' from Avinash Sajjanshetty 
Closes #3354
 2025-10-01 08:58:11 +03:00
Avinash Sajjanshetty
a360efa6e0 enable encryption feature flag by default 2025-09-30 19:04:25 +05:30
Avinash Sajjanshetty
6236dab923 mesaure read/write latencies 2025-09-29 21:58:09 +05:30
Jussi Saurio
045b11b255 bench/tpc-h: don't fail build if query 1 has output difference (known floating point precision issue) 2025-09-26 16:02:37 +03:00
Jussi Saurio
a783f82470 bench/tpch: remove "cast('yyyy-mm-dd' as datetime)" 
this causes sqlite and tursodb to interpret the value as just 'yyyy',
e.g. '1995-01-01' becomes '1995', causing a lot of the queries not to
return any results, which is not what we want.
 2025-09-26 15:55:38 +03:00
TcMits
17c91c1fe2 resolve conflicts 2025-09-22 16:03:52 +07:00
Avinash Sajjanshetty
0597ea722a Add encryption throughput test 2025-09-20 21:56:58 +05:30
Pere Diaz Bou
ff3c79d5d7 remove mvvmode and set logical log as default 2025-09-18 18:22:25 +02:00
Pere Diaz Bou
0e5b0fe8c4 perf/throughput/turso: add io option 2025-09-18 18:22:06 +02:00
Pere Diaz Bou
ba798076a0 perf/throughput/turso: add env-filter 2025-09-18 18:22:06 +02:00
Pere Diaz Bou
91c04133e9 perf/throughput/turso: allow logical log benchmark 2025-09-18 18:22:06 +02:00
Pere Diaz Bou
9b35c6efb6 rust/lib: expose mvccmode 2025-09-18 18:21:21 +02:00
Pere Diaz Bou
de8a975a0b core/mvcc: introduce MvccMode Logical Log 2025-09-18 18:21:04 +02:00
TcMits
96e4c5d241 fix issue 3144 2025-09-16 18:39:45 +07:00
pedrocarlo
3c91ae206b move as many dependencies as possible to workspace to avoid multiple versions of the same dependency 2025-09-15 17:19:36 -03:00
Jussi Saurio
26c0d72c25 perf/thrpt: add tracing 2025-09-15 14:25:18 +03:00
Pekka Enberg
dd06d2eb99 Merge 'add perf/throughput/rusqlite to workspace' from Pedro Muniz 
Closes #3116
 2025-09-15 14:24:13 +03:00
Pekka Enberg
877b28bcb3 perf/throughput/turso: Use 30 second busy timeout like in rusqlite 2025-09-15 13:57:58 +03:00
pedrocarlo
bd5dcd8d3c add timeout flag to throughput benchmark 2025-09-15 02:20:32 -03:00
pedrocarlo
3d265489dc modify semantics of busy_timeout to be more on par with sqlite 2025-09-15 02:20:32 -03:00
pedrocarlo
0586b75fbe expose function to set busy timeout duration 2025-09-15 02:20:32 -03:00
pedrocarlo
16e79ed508 slight adjustment in perf throughtput printing 2025-09-15 02:16:18 -03:00
pedrocarlo
cc48fa233b add perf/throughput/rusqlite to workspace 2025-09-14 18:29:03 -03:00
pedrocarlo
01a99f84a6 add perf/throughput/turso to workspace 2025-09-14 16:19:34 -03:00
Pekka Enberg
76a3c20a48 Merge 'perf/throughput/turso: Don't use spawn_blocking()' from Pekka Enberg 
Let's just use normal spawn()

Reviewed-by: Preston Thorpe <preston@turso.tech>

Closes #3096
 2025-09-14 17:55:44 +03:00
Pekka Enberg
a5f1bdbc9d Merge 'perf/throughput/turso: Don't interleave concurrent transactions in on…' from Pekka Enberg 
…e connection
With BEGIN CONCURRENT, we cannot use the same connection all the time
because there's no way for the transaction manager to know which
transaction belongs to what "session" -- they're all individual
statements executed in the context of the one connection.
Fixes #3093

Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>

Closes #3094
 2025-09-14 11:33:20 +03:00
Pekka Enberg
ed8b02d83e perf/throughput/turso: Don't use spawn_blocking() 
Let's just use normal spawn()
 2025-09-14 11:32:17 +03:00
Pekka Enberg
6e9ba4c561 perf/throughput/turso: Don't interleave concurrent transactions in one connection 
With BEGIN CONCURRENT, we cannot use the same connection all the time
because there's no way for the transaction manager to know which
transaction belongs to what "session" -- they're all individual
statements executed in the context of the one connection.

Fixes #3093
 2025-09-14 10:52:56 +03:00
Pekka Enberg
98af9386e2 perf/throughput/turso: Don't execute futures serially 
Looping through the futures means we're essentially executing them
serially.
 2025-09-14 10:07:21 +03:00
Pekka Enberg
0fad30a30d perf/throughput/turso: Async transactions with concurrent mode 
With `BEGIN CONCURRENT`, we should also take advantage of async
transaction processing to maximize concurrency.
 2025-09-13 13:25:16 +03:00
Pekka Enberg
8dc2e738a4 core/throughput: Add per transaction think time support 2025-09-13 13:02:43 +03:00
Pekka Enberg
964dd0cd43 perf: Add simple throughput benchmark 
This adds a simple throughput benchmark for rusqlite and Turso, allowing
to compare the two, but also MVCC and SQLite transactions.
 2025-09-12 07:35:57 +03:00
Jussi Saurio
e65742e5ff Fail CI if tursodb output differs from sqlite in tpc-h queries 2025-08-26 11:30:37 +03:00
Jussi Saurio
87bf488bbc chore: use rusqlite 0.37 with bundled sqlite everywhere 2025-08-11 15:13:57 +03:00
Jussi Saurio
51f2aad7a4 perf/clickbench: enable rest of queries since we support DISTINCT and REGEXP_REPLACE 2025-08-11 11:32:37 +03:00
Jussi Saurio
86b1232268 chore: enable indexes by default 2025-08-01 15:44:56 +03:00
Glauber Costa
5a66ed8433 measure only the time it takes to open the actual connection 
The current code includes creating the database object, which is slow.

Unfortunately the same cannot be done on the standard SQLite.
 2025-07-23 08:41:36 -05:00
Jussi Saurio
78320464ce test/clickbench: fix clickbench 
- enable indexes, otherwise it doesnt work at all
- rename limbo to tursodb
- use mode=list instead of mode=pretty for less bulky tursodb results
 2025-07-21 12:00:04 +03:00
Jussi Saurio
cc47bfba02 CSV import fixes 
- Fix not being able to create table while importing
    * The behavior now aligns with SQLite so that if the table already
      exists, all the rows are treated as data. If the table doesn't exist,
      the first row is treated as the header from which column names for the
      new table are populated.
- Insert in batches instead of one at a time
 2025-07-15 16:44:11 +03:00
Jussi Saurio
f312227825 uv run ruff format && uv run ruff check --fix 2025-07-09 10:06:29 +03:00
Pekka Enberg
b895381ae6 Revert "Merge 'Reachable assertions in Antithesis Python Test for better logging' from Pedro Muniz" 
This reverts commit dbbc3f5190, reversing
changes made to 1cd5a49705. We're missing
some mandatory parameters, causing these to fail under Antithesis.
 2025-07-08 17:51:12 +03:00
pedrocarlo
e9361c0eba add more logging to antithesis tests 
format python tests
 2025-07-07 19:11:55 -03:00
Glauber Costa
471a95bd96 add a benchmark for connection time versus number of tables 
SQLite performs poorly for connections as the number of tables increase.

Do we perform better? the same? worse?

Right now the answer is worse. Add a benchmark to help us compare.
 2025-07-01 20:10:21 -05:00