1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
|
/*
* Licensed to Elasticsearch under one or more contributor
* license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright
* ownership. Elasticsearch licenses this file to you under
* the Apache License, Version 2.0 (the "License"); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing,
* software distributed under the License is distributed on an
* "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
* KIND, either express or implied. See the License for the
* specific language governing permissions and limitations
* under the License.
*/
package org.elasticsearch.cache.recycler;
import org.elasticsearch.common.Strings;
import org.elasticsearch.common.component.AbstractComponent;
import org.elasticsearch.common.inject.Inject;
import org.elasticsearch.common.lease.Releasable;
import org.elasticsearch.common.recycler.AbstractRecyclerC;
import org.elasticsearch.common.recycler.Recycler;
import org.elasticsearch.common.settings.Setting;
import org.elasticsearch.common.settings.Settings;
import org.elasticsearch.common.unit.ByteSizeValue;
import org.elasticsearch.common.util.BigArrays;
import org.elasticsearch.common.util.concurrent.EsExecutors;
import org.elasticsearch.threadpool.ThreadPool;
import java.util.Arrays;
import java.util.Locale;
import static org.elasticsearch.common.recycler.Recyclers.concurrent;
import static org.elasticsearch.common.recycler.Recyclers.concurrentDeque;
import static org.elasticsearch.common.recycler.Recyclers.dequeFactory;
import static org.elasticsearch.common.recycler.Recyclers.none;
/** A recycler of fixed-size pages. */
public class PageCacheRecycler extends AbstractComponent implements Releasable {
public static final Setting<Type> TYPE_SETTING = new Setting<>("cache.recycler.page.type", Type.CONCURRENT.name(), Type::parse, false, Setting.Scope.CLUSTER);
public static final Setting<ByteSizeValue> LIMIT_HEAP_SETTING = Setting.byteSizeSetting("cache.recycler.page.limit.heap", "10%", false, Setting.Scope.CLUSTER);
public static final Setting<Double> WEIGHT_BYTES_SETTING = Setting.doubleSetting("cache.recycler.page.weight.bytes", 1d, 0d, false, Setting.Scope.CLUSTER);
public static final Setting<Double> WEIGHT_LONG_SETTING = Setting.doubleSetting("cache.recycler.page.weight.longs", 1d, 0d, false, Setting.Scope.CLUSTER);
public static final Setting<Double> WEIGHT_INT_SETTING = Setting.doubleSetting("cache.recycler.page.weight.ints", 1d, 0d, false, Setting.Scope.CLUSTER);
// object pages are less useful to us so we give them a lower weight by default
public static final Setting<Double> WEIGHT_OBJECTS_SETTING = Setting.doubleSetting("cache.recycler.page.weight.objects", 0.1d, 0d, false, Setting.Scope.CLUSTER);
private final Recycler<byte[]> bytePage;
private final Recycler<int[]> intPage;
private final Recycler<long[]> longPage;
private final Recycler<Object[]> objectPage;
@Override
public void close() {
bytePage.close();
intPage.close();
longPage.close();
objectPage.close();
}
private static int maximumSearchThreadPoolSize(ThreadPool threadPool, Settings settings) {
ThreadPool.Info searchThreadPool = threadPool.info(ThreadPool.Names.SEARCH);
assert searchThreadPool != null;
final int maxSize = searchThreadPool.getMax();
if (maxSize <= 0) {
// happens with cached thread pools, let's assume there are at most 2x ${number of processors} threads
return 2 * EsExecutors.boundedNumberOfProcessors(settings);
} else {
return maxSize;
}
}
@Inject
public PageCacheRecycler(Settings settings, ThreadPool threadPool) {
super(settings);
final Type type = TYPE_SETTING .get(settings);
final long limit = LIMIT_HEAP_SETTING .get(settings).bytes();
final int availableProcessors = EsExecutors.boundedNumberOfProcessors(settings);
final int searchThreadPoolSize = maximumSearchThreadPoolSize(threadPool, settings);
// We have a global amount of memory that we need to divide across data types.
// Since some types are more useful than other ones we give them different weights.
// Trying to store all of them in a single stack would be problematic because eg.
// a work load could fill the recycler with only byte[] pages and then another
// workload that would work with double[] pages couldn't recycle them because there
// is no space left in the stack/queue. LRU/LFU policies are not an option either
// because they would make obtain/release too costly: we really need constant-time
// operations.
// Ultimately a better solution would be to only store one kind of data and have the
// ability to intepret it either as a source of bytes, doubles, longs, etc. eg. thanks
// to direct ByteBuffers or sun.misc.Unsafe on a byte[] but this would have other issues
// that would need to be addressed such as garbage collection of native memory or safety
// of Unsafe writes.
final double bytesWeight = WEIGHT_BYTES_SETTING .get(settings);
final double intsWeight = WEIGHT_INT_SETTING .get(settings);
final double longsWeight = WEIGHT_LONG_SETTING .get(settings);
final double objectsWeight = WEIGHT_OBJECTS_SETTING .get(settings);
final double totalWeight = bytesWeight + intsWeight + longsWeight + objectsWeight;
final int maxPageCount = (int) Math.min(Integer.MAX_VALUE, limit / BigArrays.PAGE_SIZE_IN_BYTES);
final int maxBytePageCount = (int) (bytesWeight * maxPageCount / totalWeight);
bytePage = build(type, maxBytePageCount, searchThreadPoolSize, availableProcessors, new AbstractRecyclerC<byte[]>() {
@Override
public byte[] newInstance(int sizing) {
return new byte[BigArrays.BYTE_PAGE_SIZE];
}
@Override
public void recycle(byte[] value) {
// nothing to do
}
});
final int maxIntPageCount = (int) (intsWeight * maxPageCount / totalWeight);
intPage = build(type, maxIntPageCount, searchThreadPoolSize, availableProcessors, new AbstractRecyclerC<int[]>() {
@Override
public int[] newInstance(int sizing) {
return new int[BigArrays.INT_PAGE_SIZE];
}
@Override
public void recycle(int[] value) {
// nothing to do
}
});
final int maxLongPageCount = (int) (longsWeight * maxPageCount / totalWeight);
longPage = build(type, maxLongPageCount, searchThreadPoolSize, availableProcessors, new AbstractRecyclerC<long[]>() {
@Override
public long[] newInstance(int sizing) {
return new long[BigArrays.LONG_PAGE_SIZE];
}
@Override
public void recycle(long[] value) {
// nothing to do
}
});
final int maxObjectPageCount = (int) (objectsWeight * maxPageCount / totalWeight);
objectPage = build(type, maxObjectPageCount, searchThreadPoolSize, availableProcessors, new AbstractRecyclerC<Object[]>() {
@Override
public Object[] newInstance(int sizing) {
return new Object[BigArrays.OBJECT_PAGE_SIZE];
}
@Override
public void recycle(Object[] value) {
Arrays.fill(value, null); // we need to remove the strong refs on the objects stored in the array
}
});
assert BigArrays.PAGE_SIZE_IN_BYTES * (maxBytePageCount + maxIntPageCount + maxLongPageCount + maxObjectPageCount) <= limit;
}
public Recycler.V<byte[]> bytePage(boolean clear) {
final Recycler.V<byte[]> v = bytePage.obtain();
if (v.isRecycled() && clear) {
Arrays.fill(v.v(), (byte) 0);
}
return v;
}
public Recycler.V<int[]> intPage(boolean clear) {
final Recycler.V<int[]> v = intPage.obtain();
if (v.isRecycled() && clear) {
Arrays.fill(v.v(), 0);
}
return v;
}
public Recycler.V<long[]> longPage(boolean clear) {
final Recycler.V<long[]> v = longPage.obtain();
if (v.isRecycled() && clear) {
Arrays.fill(v.v(), 0L);
}
return v;
}
public Recycler.V<Object[]> objectPage() {
// object pages are cleared on release anyway
return objectPage.obtain();
}
private static <T> Recycler<T> build(Type type, int limit, int estimatedThreadPoolSize, int availableProcessors, Recycler.C<T> c) {
final Recycler<T> recycler;
if (limit == 0) {
recycler = none(c);
} else {
recycler = type.build(c, limit, estimatedThreadPoolSize, availableProcessors);
}
return recycler;
}
public enum Type {
QUEUE {
@Override
<T> Recycler<T> build(Recycler.C<T> c, int limit, int estimatedThreadPoolSize, int availableProcessors) {
return concurrentDeque(c, limit);
}
},
CONCURRENT {
@Override
<T> Recycler<T> build(Recycler.C<T> c, int limit, int estimatedThreadPoolSize, int availableProcessors) {
return concurrent(dequeFactory(c, limit / availableProcessors), availableProcessors);
}
},
NONE {
@Override
<T> Recycler<T> build(Recycler.C<T> c, int limit, int estimatedThreadPoolSize, int availableProcessors) {
return none(c);
}
};
public static Type parse(String type) {
try {
return Type.valueOf(type.toUpperCase(Locale.ROOT));
} catch (IllegalArgumentException e) {
throw new IllegalArgumentException("no type support [" + type + "]");
}
}
abstract <T> Recycler<T> build(Recycler.C<T> c, int limit, int estimatedThreadPoolSize, int availableProcessors);
}
}
|
