FastUtil Concurrent Wrapper

Description

Set of concurrent wrappers around fastutil primitive maps.

How we use it at trivago.

Main purpose is to provide useful concurrent builders around fastutil primitive maps with flexible locking policy.

Advantages over java.util wrappers:

• builders provide maps with buckets,
• every map uses striped ReadWriteLocks instead of synchronized(mutex); one RW-lock per map's bucket,
• two lock modes: standard and busy-waiting (could be good for low-latency systems),
• no extra memory on stack -- API based on primitive types.

Check usage section for more details.

Note: currently the lib contains wrappers not for every primitive map. Feel free to contribute.

Install

Maven

<dependency>
    <groupId>com.trivago</groupId>
    <artifactId>fastutil-concurrent-wrapper</artifactId>
    <version>0.2.2</version>
</dependency>

Gradle

implementation group: 'com.trivago', name: 'fastutil-concurrent-wrapper', version: '0.2.2'

Usage

Builder options

• number of buckets -- number of buckets in the map (default 8),
• default value -- default value, for getOrDefault() method
• initial capacity -- initial map capacity (default 100_000),
• concurrent mode -- lock mode: default and busy-waiting,
• load factor -- map load factor (default 0.8f).

Basic usage

ConcurrentLongLongMapBuilder b = ConcurrentLongLongMapBuilder.newBuilder()
        .withBuckets(2)
        .withDefaultValue(0)
        .withInitialCapacity(100)
        .withMode(ConcurrentLongLongMapBuilder.MapMode.BUSY_WAITING)
        .withLoadFactor(0.9f);

LongLongMap map = b.build();

map.put(1L, 10L);
long v = map.get(1L);

Examples of creation and usage could be found inside test directory;

MapMode

Currently, we offer two locking modes:

• blocking (default),
• busy-waiting.

JMH tests

For running JMH tests just execute:

./gradlew jmh

Results for FastutilWrapper BusyWaiting mode vs FastutilWrapper Default mode vs java.util wrappers

Throughput (more is better)

Benchmark                                                        Mode  Cnt         Score         Error  Units

FastutilWrapperBusyWaitingBenchmark.testRandomAllOpsThroughput  thrpt   15  14517457,055 ?  795637,784  ops/s
FastutilWrapperBusyWaitingBenchmark.testRandomGetThroughput     thrpt   15  16610181,320 ? 1456776,589  ops/s
FastutilWrapperBusyWaitingBenchmark.testRandomPutThroughput     thrpt   13  11706178,916 ? 2547333,524  ops/s

FastutilWrapperDefaultBenchmark.testRandomAllOpsThroughput      thrpt   15   7385357,514 ? 1127356,032  ops/s
FastutilWrapperDefaultBenchmark.testRandomGetThroughput         thrpt   15  16190621,923 ? 1836415,022  ops/s
FastutilWrapperDefaultBenchmark.testRandomPutThroughput         thrpt   15   8945369,395 ? 1225460,217  ops/s

JavaUtilWrapperBenchmark.testRandomAllOpsThroughput             thrpt   15   4921201,916 ?  410471,239  ops/s
JavaUtilWrapperBenchmark.testRandomGetThroughput                thrpt   15   7827123,690 ?  557193,670  ops/s
JavaUtilWrapperBenchmark.testRandomPutThroughput                thrpt   15   4832517,371 ? 1122344,647  ops/s

AverageTime per ops (less is better)

Benchmark                                                        Mode  Cnt         Score         Error  Units

FastutilWrapperBusyWaitingBenchmark.testRandomAllOpsAvgTime      avgt   15       268,790 ?      22,526  ns/op
FastutilWrapperBusyWaitingBenchmark.testRandomGetAvgTime         avgt   15       231,552 ?      16,116  ns/op
FastutilWrapperBusyWaitingBenchmark.testRandomPutAvgTime         avgt   10       292,246 ?      49,757  ns/op

FastutilWrapperDefaultBenchmark.testRandomAllOpsAvgTime          avgt   15       467,381 ?       9,790  ns/op
FastutilWrapperDefaultBenchmark.testRandomGetAvgTime             avgt   15       237,683 ?      14,167  ns/op
FastutilWrapperDefaultBenchmark.testRandomPutAvgTime             avgt   15       427,441 ?      25,116  ns/op

JavaUtilWrapperBenchmark.testRandomAllOpsAvgTime                 avgt   15       781,869 ?     191,081  ns/op
JavaUtilWrapperBenchmark.testRandomGetAvgTime                    avgt   15       470,869 ?      33,198  ns/op
JavaUtilWrapperBenchmark.testRandomPutAvgTime                    avgt   15       964,613 ?     422,648  ns/op

The machine

MacBook Pro (15-inch, 2019)
Processor 2,6 GHz 6-Core Intel Core i7
Memory 16 GB 2400 MHz DDR4

Maintainers

A-Z surname order

• @mchernyakov
• @erdoganf
• @sarveswaran-m