{ "date": "2026-10-16", "issue": 237, "source_commit": "dd53a0316130e280b4b0b7605e8ddb404f76d15d", "variant": "STRING cache with copying Arrow reader probes, before borrowed probe experiment", "flink": "2.2.1", "jdk": "17", "cpu": "Intel i7-12650H", "configuration": { "heap": "2 GiB", "parallelism": 1, "build": "release mimalloc", "production_ipc": true, "both_transposes": true, "batch_rows": 1024, "warmup_pairs": 2, "measured_pairs": 5, "metrics_cluster": "fresh per job", "boundary": "executeSql().await including source/sink, state and conversions" }, "measurements": [ { "label": "string-cache-16", "rows": 2000000, "keys": 16, "stock_trials_s": [ 0.727575729, 0.708105867, 0.717205442, 0.760919697, 0.714603522 ], "native_trials_s": [ 0.729680903, 0.714809868, 0.716069469, 0.691952305, 0.622429621 ], "stock_median_s": 0.717205442, "native_median_s": 0.714809868, "stock_range_s": [ 0.708105867, 0.760919697 ], "native_range_s": [ 0.622429621, 0.729680903 ], "activity": { "exit_code": 0, "external_activity": [], "interpretation": "Any external build activity disqualifies this run from headline comparisons." }, "raw_log": "jvm-aggregate-copying-string-probe-string-cache-16-2026-10-16.log.gz", "command": [ "mvn", "-o", "-B", "-ntp", "surefire:test@default-test", "-Pbench", "-pl", "streamfusion-runtime", "-Dtest=JvmAggregateBenchmark", "-Dnative.build.skip=true", "-Dnative.resource.directory=/native/target/232-json-fixed", "-Dsf.benchmarkJvmArgs=-Xmx2g -Dstreamfusion.exchange.zeroCopyLocal=false", "-Dsf.extraJvmArgs=-Djava.library.path=/native/target/232-json-fixed -Dstreamfusion.experimental.jvmGroupAggregates=true -Dstreamfusion.experimental.jvmAccumulatorCacheEntries=65536 -Xlog:gc*,safepoint:file=/native/target/237-jvm-aggregate/string-cache-16-gc.log:time,uptime,level,tags", "-Djvmagg.expectNative=true", "-Djvmagg.cases=udaf", "-Djvmagg.rows=2000000", "-Djvmagg.keys=16", "-Djvmagg.warmup=2", "-Djvmagg.runs=5", "-Djvmagg.keyType=string", "-Djvmagg.expectCache=true" ] }, { "label": "string-cache-65k", "rows": 2000000, "keys": 65536, "stock_trials_s": [ 1.305394494, 1.283711585, 1.876215893, 1.210459842, 1.173259626 ], "native_trials_s": [ 1.449921878, 1.427949061, 1.315862028, 1.405219242, 1.18262709 ], "stock_median_s": 1.283711585, "native_median_s": 1.405219242, "stock_range_s": [ 1.173259626, 1.876215893 ], "native_range_s": [ 1.18262709, 1.449921878 ], "activity": { "exit_code": 0, "external_activity": [], "interpretation": "Any external build activity disqualifies this run from headline comparisons." }, "raw_log": "jvm-aggregate-copying-string-probe-string-cache-65k-2026-10-16.log.gz", "command": [ "mvn", "-o", "-B", "-ntp", "surefire:test@default-test", "-Pbench", "-pl", "streamfusion-runtime", "-Dtest=JvmAggregateBenchmark", "-Dnative.build.skip=true", "-Dnative.resource.directory=/native/target/232-json-fixed", "-Dsf.benchmarkJvmArgs=-Xmx2g -Dstreamfusion.exchange.zeroCopyLocal=false", "-Dsf.extraJvmArgs=-Djava.library.path=/native/target/232-json-fixed -Dstreamfusion.experimental.jvmGroupAggregates=true -Dstreamfusion.experimental.jvmAccumulatorCacheEntries=65536 -Xlog:gc*,safepoint:file=/native/target/237-jvm-aggregate/string-cache-65k-gc.log:time,uptime,level,tags", "-Djvmagg.expectNative=true", "-Djvmagg.cases=udaf", "-Djvmagg.rows=2000000", "-Djvmagg.keys=65536", "-Djvmagg.warmup=2", "-Djvmagg.runs=5", "-Djvmagg.keyType=string", "-Djvmagg.expectCache=true" ] }, { "label": "string-cache-distinct", "rows": 2000000, "keys": 2000000, "stock_trials_s": [ 1.988491944, 2.004538852, 1.972298231, 2.018344024, 1.955629815 ], "native_trials_s": [ 2.159814393, 2.081507494, 2.203838157, 2.078862545, 2.071798086 ], "stock_median_s": 1.988491944, "native_median_s": 2.081507494, "stock_range_s": [ 1.955629815, 2.018344024 ], "native_range_s": [ 2.071798086, 2.203838157 ], "activity": { "exit_code": 0, "external_activity": [], "interpretation": "Any external build activity disqualifies this run from headline comparisons." }, "raw_log": "jvm-aggregate-copying-string-probe-string-cache-distinct-2026-10-16.log.gz", "command": [ "mvn", "-o", "-B", "-ntp", "surefire:test@default-test", "-Pbench", "-pl", "streamfusion-runtime", "-Dtest=JvmAggregateBenchmark", "-Dnative.build.skip=true", "-Dnative.resource.directory=/native/target/232-json-fixed", "-Dsf.benchmarkJvmArgs=-Xmx2g -Dstreamfusion.exchange.zeroCopyLocal=false", "-Dsf.extraJvmArgs=-Djava.library.path=/native/target/232-json-fixed -Dstreamfusion.experimental.jvmGroupAggregates=true -Dstreamfusion.experimental.jvmAccumulatorCacheEntries=65536 -Xlog:gc*,safepoint:file=/native/target/237-jvm-aggregate/string-cache-distinct-gc.log:time,uptime,level,tags", "-Djvmagg.expectNative=true", "-Djvmagg.cases=udaf", "-Djvmagg.rows=2000000", "-Djvmagg.keys=2000000", "-Djvmagg.warmup=2", "-Djvmagg.runs=5", "-Djvmagg.keyType=string", "-Djvmagg.expectCache=true" ] } ], "profile": { "rows": 1000000, "keys": 32768, "warmup_pairs": 2, "pairs": 3, "includes": [ "JVM UDAF", "native builtin SUM Rust control" ], "timings_excluded": true, "activity": { "exit_code": 0, "external_activity": [], "timing_excluded": true }, "stacks": { "cpu": { "file": "jvm-aggregate-copying-string-probe-cpu-2026-10-16.collapsed.gz", "sha256": "126613f8b1e1e92d58b7b637f2220018047c93fbcda54409efc6bba3a6caadfd", "units": "CPU samples", "inclusive_jvm_aggregate_total": 1383, "inclusive_methods": { "JvmAccumulatorCache": 736, "BinaryRowDataKeySelector.getKey": 6, "HashMap.getNode": 582, "ColumnarRowData.getString": 85, "RowDataSerializer.copy": 140 }, "leading_jvm_leaves": [ [ "java/util/HashMap.getNode_[i]", 349 ], [ "org/apache/flink/core/memory/MemorySegment.equalTo_[i]", 100 ], [ "org/apache/flink/table/data/RowData.lambda$createFieldGetter$245ca7d1$7_[i]", 62 ], [ "org/apache/flink/table/runtime/operators/aggregate/GroupAggFunction.processElement_[j]", 51 ], [ "tech/streamfusion/operator/JvmAccumulatorCache$CachedState.keyFor_[i]", 28 ], [ "org/apache/flink/table/data/binary/BinarySegmentUtils.inFirstSegment_[i]", 25 ], [ "org/apache/flink/table/data/binary/MurmurHashUtils.mixH1_[i]", 23 ], [ "tech/streamfusion/PortableSqlFixtures$Sum.getValue_[i]", 23 ], [ "org/apache/flink/table/data/binary/LazyBinaryFormat.ensureMaterialized_[j]", 22 ], [ "org/apache/flink/table/data/binary/BinaryStringData.equals_[j]", 22 ], [ "jdk/internal/misc/Unsafe.copyMemory_[i]", 21 ], [ "tech/streamfusion/operator/JvmAccumulatorCache$CachedState.update_[i]", 21 ] ], "leading_rust_frames": [ [ "streamfusion::group_agg::GroupAggregator::update", 614 ], [ "streamfusion::group_agg::output_of", 80 ], [ "streamfusion::group_agg::GroupChanges::push", 35 ], [ "streamfusion::exchange::coalesce::coalesce_batches", 22 ], [ "streamfusion::flink_key::write_value", 20 ], [ "streamfusion::aggregates::scalars_to_array", 13 ], [ "streamfusion::flink_key::BinaryRowWriter::write_variable", 10 ], [ "core::ptr::drop_in_place>", 6 ], [ "core::ptr::drop_in_place", 6 ], [ "streamfusion::group_agg::group_agg_state_bytes", 6 ] ] }, "alloc": { "file": "jvm-aggregate-copying-string-probe-alloc-2026-10-16.collapsed.gz", "sha256": "20fc14c2a8a53a8687412f482128374c8318a5a76a3424b6e8a28cddad0208eb", "units": "sampled allocation bytes", "inclusive_jvm_aggregate_total": 4998561792, "inclusive_methods": { "JvmAccumulatorCache": 1060110336, "ColumnarRowData.getString": 1811939328, "RowDataSerializer.copy": 1891631104, "BinaryRowDataKeySelector.getKey": 62914560 }, "leading_jvm_leaves": [ [ "org.apache.flink.core.memory.MemorySegment_[i]", 809500672 ], [ "java.lang.Long_[i]", 612368384 ], [ "org.apache.flink.table.data.GenericRowData_[i]", 608174080 ], [ "java.lang.Object[]_[i]", 577765376 ], [ "byte[]_[i]", 572522496 ], [ "org.apache.flink.table.data.binary.BinaryStringData_[i]", 460324864 ], [ "org.apache.flink.table.data.binary.BinarySection_[i]", 445644800 ], [ "org.apache.flink.core.memory.MemorySegment[]_[i]", 369098752 ], [ "java.util.concurrent.atomic.AtomicBoolean_[i]", 205520896 ], [ "tech.streamfusion.PortableSqlFixtures$SumState_[i]", 177209344 ], [ "org.apache.flink.streaming.runtime.streamrecord.StreamRecord_[i]", 121634816 ], [ "java.util.HashMap$Node_[i]", 5242880 ] ], "leading_rust_frames": [] } } }, "admitted": false, "interpretation": [ "No observed competing Java/Rust build/job work in 500 ms sampling", "16 keys roughly ties stock; medium and distinct medians regress. No default admission.", "The local profile is on different hardware/JDK from the isolated baseline; it is not a matched causal comparison. Key selector sampling is small, but cache lookup and temporary string wrappers remain concrete targets.", "Profile timings excluded. Inclusive sampled counts overlap downstream copying and are not copied-byte or peak-memory measurements." ], "preflight_limitation": "Historical v1 benchmark explains only native before timing. Stock/native admission must use corrected symmetric preflight. Profiled NativeJvmGroup stacks remain diagnostic; all profile timings excluded." }