bf3d0b3543
Observed 3% throughput improvement in sstable-heavy workload bounded by CPU.
SStable parsing involves lots of buffer operations which obtain and
destroy resource_units. Before the patch, reosurce_unit destruction
invoked maybe_admit_waiters(), which performs some computations on
waiting permits. We don't really need to admit on each change of
resources, since the CPU is used by other things anyway. We can batch
the computation. There is already a fiber which does this for
processing the _ready_list. We can reuse it for processing _wait_list
as well.
The changes violate an assumption made by tests that releasing
resources immediately triggers an admission check. Therefore, some of
the BOOST_REQUIRE_EQUAL needs to be replaced with REQUIRE_EVENTUALLY_EQUAL
as the admision check is now done in the fiber processing the _ready_list.
`perf-simple-query` --tablets --smp 1 -m 1G results obtained for
fixed 400MHz frequency:
Before:
```
enable-cache=1
Running test with config: {partitions=10000, concurrency=100, mode=read, frontend=cql, query_single_key=no, counters=no}
Disabling auto compaction
Creating 10000 partitions...
112590.60 tps ( 63.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 41353 insns/op, 17992 cycles/op, 0 errors)
122620.68 tps ( 63.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 41310 insns/op, 17713 cycles/op, 0 errors)
118169.48 tps ( 63.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 41353 insns/op, 17857 cycles/op, 0 errors)
120634.65 tps ( 63.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 41328 insns/op, 17733 cycles/op, 0 errors)
117317.18 tps ( 63.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 41347 insns/op, 17822 cycles/op, 0 errors)
throughput: mean=118266.52 standard-deviation=3797.81 median=118169.48 median-absolute-deviation=2368.13 maximum=122620.68 minimum=112590.60
instructions_per_op: mean=41337.86 standard-deviation=18.73 median=41346.89 median-absolute-deviation=14.64 maximum=41352.53 minimum=41309.83
cpu_cycles_per_op: mean=17823.50 standard-deviation=111.75 median=17821.97 median-absolute-deviation=90.45 maximum=17992.04 minimum=17713.00
```
After
```
enable-cache=1
Running test with config: {partitions=10000, concurrency=100, mode=read, frontend=cql, query_single_key=no, counters=no}
Disabling auto compaction
Creating 10000 partitions...
123689.63 tps ( 63.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 40997 insns/op, 17384 cycles/op, 0 errors)
129643.24 tps ( 63.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 40997 insns/op, 17325 cycles/op, 0 errors)
128907.27 tps ( 63.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 41009 insns/op, 17325 cycles/op, 0 errors)
130342.56 tps ( 63.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 40993 insns/op, 17286 cycles/op, 0 errors)
130294.09 tps ( 63.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 40972 insns/op, 17336 cycles/op, 0 errors)
throughput: mean=128575.36 standard-deviation=2792.75 median=129643.24 median-absolute-deviation=1718.73 maximum=130342.56 minimum=123689.63
instructions_per_op: mean=40993.51 standard-deviation=13.23 median=40996.73 median-absolute-deviation=3.30 maximum=41008.86 minimum=40972.48
cpu_cycles_per_op: mean=17331.16 standard-deviation=35.02 median=17324.84 median-absolute-deviation=6.49 maximum=17383.97 minimum=17286.33
```
Closes scylladb/scylladb#21918
[avi: patch was co-authored by Łukasz Paszkowski <lukasz.paszkowski@scylladb.com>]
Scylla in-source tests.
For details on how to run the tests, see docs/dev/testing.md
Shared C++ utils, libraries are in lib/, for Python - pylib/
alternator - Python tests which connect to a single server and use the DynamoDB API unit, boost, raft - unit tests in C++ cqlpy - Python tests which connect to a single server and use CQL topology* - tests that set up clusters and add/remove nodes cql - approval tests that use CQL and pre-recorded output rest_api - tests for Scylla REST API Port 9000 scylla-gdb - tests for scylla-gdb.py helper script nodetool - tests for C++ implementation of nodetool
If you can use an existing folder, consider adding your test to it. New folders should be used for new large categories/subsystems, or when the test environment is significantly different from some existing suite, e.g. you plan to start scylladb with different configuration, and you intend to add many tests and would like them to reuse an existing Scylla cluster (clusters can be reused for tests within the same folder).
To add a new folder, create a new directory, and then
copy & edit its suite.ini.