mirrors/scylladb
1
0
Fork 0
mirror of https://github.com/scylladb/scylladb.git synced 2026-04-29 04:37:00 +00:00
Code Issues Packages Projects Releases Wiki Activity
26,515 Commits 241 Branches 541 Tags
fdfcda97d70393c8f40efcb4fc64df233ebf49cc
Commit Graph

15 Commits

Author SHA1 Message Date
Piotr Sarna
6c6ccda8a0 perf: add alternator frontend to perf_simple_query 
The perf_simple_query tool is extended with another protocol
aside from CQL - alternator. The alternative (pun intended) benchmark
can be executed by using the `--alternator X` parameter, where X
specifies one of the alternator's mandatory write isolation options:
 - "forbid_rmw" - forbids RMW (read-modify-write) requests
 - "unsafe" - never uses LWT (lightweight transactions), even for RMW
 - "always_use_lwt" - uses LWT even for non-RMW requests
 - "only_rmw_uses_lwt" - that one's rather self-explanatory

Alternator cooperates with existing --write and --delete parameters.

Aside from being able to check for improvements/regressions
in the alternator module, it's also possible to check how different
isolation levels influence the number of allocations and overall
performance, or to compare alternator against CQL.

$ ./build/release/test/perf/perf_simple_query_g --smp 1 \
    --write --alternator only_rmw_uses_lwt --default-log-level error
random-seed=1235000092
Started alternator executor
10873.76 tps (202.9 allocs/op,  12.4 tasks/op,  369921 insns/op)
11096.09 tps (202.7 allocs/op,  12.1 tasks/op,  374792 insns/op)
11100.09 tps (203.0 allocs/op,  12.1 tasks/op,  376469 insns/op)
11068.98 tps (203.1 allocs/op,  12.1 tasks/op,  377132 insns/op)
11081.24 tps (203.2 allocs/op,  12.1 tasks/op,  377290 insns/op)

median 11081.24 tps (203.2 allocs/op,  12.1 tasks/op,  377290 insns/op)
median absolute deviation: 14.85
maximum: 11100.09
minimum: 10873.76

$ ./build/release/test/perf/perf_simple_query_g --smp 1 \
    --random-seed 1235000092 --write --alternator always_use_lwt \
    --default-log-level error
random-seed=1235000092
Started alternator executor
3605.35 tps (877.4 allocs/op, 174.6 tasks/op,  986666 insns/op)
3555.71 tps (890.0 allocs/op, 174.4 tasks/op, 1006945 insns/op)
3530.20 tps (899.7 allocs/op, 174.1 tasks/op, 1021908 insns/op)
3437.65 tps (908.2 allocs/op, 174.6 tasks/op, 1033992 insns/op)
3409.88 tps (913.2 allocs/op, 174.4 tasks/op, 1041240 insns/op)

median 3530.20 tps (899.7 allocs/op, 174.1 tasks/op, 1021908 insns/op)
median absolute deviation: 75.15
maximum: 3605.35
minimum: 3409.88
 2021-05-18 15:10:31 +02:00
Botond Dénes
dca808dd51 perf/perf_simple_query: add --enable-cache option 
Allowing for testing performance with/out cache.

Signed-off-by: Botond Dénes <bdenes@scylladb.com>
Message-Id: <20210517045402.16153-1-bdenes@scylladb.com>
 2021-05-17 14:06:18 +02:00
Avi Kivity
daeddda7cc treewide: remove inclusions of storage_proxy.hh from headers 
storage_proxy.hh is huge and includes many headers itself, so
remove its inclusions from headers and re-add smaller headers
where needed (and storage_proxy.hh itself in source files that
need it).

Ref #1.
 2021-04-20 21:23:00 +03:00
Benny Halevy
9c89702fb2 perf_simple_query: use tests::random::get_int for reproducible results 
Support for random-seed was added in 4ad06c7eeb
but the program still uses std::rand() to draw random keys.

Use tests::random::get_int instead so we can get reprodicible
sequence of keys given a particular random-seed.

Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Message-Id: <20210418104455.82086-1-bhalevy@scylladb.com>
 2021-04-19 17:22:53 +02:00
Avi Kivity
202c631dee test: perf: perf_simple_query: collect allocation and task statistics 
Calculate and display the number of memory allocations and tasks
executed per operation. Sample results (--smp 1):

180022.46 tps (90 allocs/op, 20 tasks/op)
178963.44 tps (90 allocs/op, 20 tasks/op)
178702.41 tps (90 allocs/op, 20 tasks/op)
177679.74 tps (90 allocs/op, 20 tasks/op)
179539.36 tps (90 allocs/op, 20 tasks/op)

median 178963.44 tps (90 allocs/op, 20 tasks/op)
median absolute deviation: 575.92
maximum: 180022.46
minimum: 177679.74

This allows less noisy tracking of how some changes impact performance.
 2021-04-07 17:54:48 +03:00
Avi Kivity
37b41d7764 test: add missing #include <fstream> 
std::ofstream is used, but there is no direct include for it. This
fails the build with libstdc++ 11.

Closes #8050
 2021-02-09 14:45:20 +02:00
Piotr Jastrzebski
c001374636 codebase wide: replace count with contains 
C++20 introduced `contains` member functions for maps and sets for
checking whether an element is present in the collection. Previously
`count` function was often used in various ways.

`contains` does not only express the intend of the code better but also
does it in more unified way.

This commit replaces all the occurences of the `count` with the
`contains`.

Tests: unit(dev)

Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
Message-Id: <b4ef3b4bc24f49abe04a2aba0ddd946009c9fcb2.1597314640.git.piotr@scylladb.com>
 2020-08-15 20:26:02 +03:00
Pavel Emelyanov
ef181fb2d0 test: Add option to flush memtables for perf_simple_query 
The test in question measures the speed of memtables, not
the row_cache. With this option it can do both.

Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Message-Id: <20200507140603.12350-1-xemul@scylladb.com>
 2020-05-07 16:09:40 +02:00
Piotr Sarna
bec95a0605 treewide: use thread-safe variant of localtime 
In order to ensure thread-safety, all usages of localtime()
are replaced with localtime_r(), which may accept a local
buffer.

Tests: unit(dev)
Fixes #6364
Message-Id: <ad4a0c0e1707f0318325718715a3a647e3ebfdfe.1588592156.git.sarna@scylladb.com>
 2020-05-04 14:46:08 +03:00
Raphael S. Carvalho
5ac0d31323 test: perf_simple_query: fix test with smp count > 1 
that code doesn't run under a thread, so let's futurize it.
the code worked with single cpu because get() returns right away
due to no deferring point.

Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20200427155303.82763-1-raphaelsc@scylladb.com>
 2020-04-27 18:58:25 +03:00
Rafael Ávila de Espíndola
c5795e8199 everywhere: Replace engine().cpu_id() with this_shard_id() 
This is a bit simpler and might allow removing a few includes of
reactor.hh.

Signed-off-by: Rafael Ávila de Espíndola <espindola@scylladb.com>
Message-Id: <20200326194656.74041-1-espindola@scylladb.com>
 2020-03-27 11:40:03 +03:00
Rafael Ávila de Espíndola
2679c0cc87 perf_simple_query: Pass a string_view to make_counter_schema 
With this we don't need to construct a sstring just to call
make_counter_schema.

Signed-off-by: Rafael Ávila de Espíndola <espindola@scylladb.com>
 2020-02-28 08:36:27 -08:00
Rafael Ávila de Espíndola
17f12a8197 perf_simple_query: Call set_abort_on_internal_error(true) 
We should never ignore an internal error in a perf test.

Signed-off-by: Rafael Ávila de Espíndola <espindola@scylladb.com>
Message-Id: <20200225055745.321086-2-espindola@scylladb.com>
 2020-02-26 18:22:05 +02:00
Rafael Ávila de Espíndola
c6897dcbea perf_simple_query: Simplify with seastar::thread 
There is no reason not to use a seastar::thread in setup code.

Signed-off-by: Rafael Ávila de Espíndola <espindola@scylladb.com>
Message-Id: <20200225055745.321086-1-espindola@scylladb.com>
 2020-02-26 18:22:04 +02:00
Konstantin Osipov
1c8736f998 tests: move all test source files to their new locations 
1. Move tests to test (using singular seems to be a convention
   in the rest of the code base)
2. Move boost tests to test/boost, other
   (non-boost) unit tests to test/unit, tests which are
   expected to be run manually to test/manual.

Update configure.py and test.py with new paths to tests.
 2019-12-16 17:47:42 +03:00