問題

使用java開源項目經常需要調優jvm，以優化gc。對于gc，如果對象都是短時對象，那么jvm相對容易優化，假如碰上像solr使用自帶java cache的項目，那么gc嚴重受限于cache，因為cache對象并非短時對象，以至于young gc常常伴有大量的內存對象拷貝，嚴重影響gc性能。

Ehcache BigMemory

Java的內存管理機制極其不適用于cache，最好的辦法是使用jni實現的cache系統。另一種通用辦法：Ehcache BigMemory(http://ehcache.org/)。BigMemory extends Ehcache's' capabilities with an off-heap store that frees you from GC’s constraints.

對于BigMemory，直接下載免費的32G限制的版本(注: 每個jvm進程最多使用32G的off-heap空間，對大多數應用已足夠)。

關于如何使用，參見官方文檔： http://terracotta.org/documentation/4.0/bigmemorygo/get-started

使用示例可參考代碼中自帶的樣例：bigmemory-go-4.0.0/code-samples/src/main/java/com/bigmemory/samples

樣例代碼缺少編譯配置build.xml, 將下面的 build.xml 放在 bigmemory-go-4.0.0/code-samples 即可使用ant 編譯示例代碼：

<project name="bigmemory" basedir="."><property name="build.dir" value="${basedir}/build" /><property name="src.class.dir" value="${build.dir}" /><property name="src.dir" value="${basedir}/src" /><property name="lib.dir" value="${basedir}/../lib" /><property name="config.dir" value="${basedir}/config" /><path id="base.classpath"><pathelement location="${src.class.dir}" /><pathelement location="${config.dir}" /><fileset dir="${lib.dir}"><include name="**/*.jar" /></fileset></path><path id="classpath"><path refid="base.classpath" /><fileset dir="${lib.dir}"><include name="**/*.jar" /></fileset></path><path id="build.src.path"><pathelement location="${src.class.dir}" /></path><target name="clean" description="clean"><delete dir="${build.dir}" /></target><target name="compile" depends="clean" description="compile"><mkdir dir="${src.class.dir}" /><javac srcdir="${src.dir}" destdir="${src.class.dir}" source="1.6" debug="on" encoding="utf-8" includeantruntime="false"><classpath refid="base.classpath" /></javac></target><target name="jar" depends="compile" description="jar"><jar destfile="${build.dir}/bigmemory.jar"><fileset dir="${src.class.dir}"><exclude name="**/timer/**" /></fileset></jar></target> </project>

配置說明：bigmemory-go-4.0.0/config-samples/ehcache.xml 詳細說明了配置參數。

限制：

1、存儲對象全部使用 java.io.Serializable 做序列化和反序列化，性能有損失。

2、off-heap空間一經分配不可調整。

solr緩存

引入Ehcache bigmemory是為了優化solr的緩存。下面代碼是基于solr cache基類實現的ehcache緩存類，使用上同于solr.FastLRUCache，需要ehcache的外部配置文件。

package org.apache.solr.search;import org.apache.solr.common.SolrException; import org.apache.solr.common.util.NamedList; import org.apache.solr.common.util.SimpleOrderedMap; import org.apache.solr.core.SolrCore;import java.util.*; import java.util.concurrent.atomic.AtomicLong; import java.io.IOException; import java.net.URL;import net.sf.ehcache.Cache; import net.sf.ehcache.CacheManager; import net.sf.ehcache.Element; import net.sf.ehcache.config.CacheConfiguration; import net.sf.ehcache.config.Configuration; import net.sf.ehcache.config.MemoryUnit;/*** @version $Id: EhCacheWrapper.java 2013-03-27 zhenjing chen $*/ public class EhCacheWrapper implements SolrCache {/* An instance of this class will be shared across multiple instances* of an LRUCache at the same time. Make sure everything is thread safe.*/private static class CumulativeStats {AtomicLong lookups = new AtomicLong();AtomicLong hits = new AtomicLong();AtomicLong inserts = new AtomicLong();AtomicLong evictions = new AtomicLong();}private CumulativeStats stats;// per instance stats. The synchronization used for the map will also be// used for updating these statistics (and hence they are not AtomicLongsprivate long lookups;private long hits;private long inserts;private long evictions;private long warmupTime = 0;private CacheManager manager = null;private Cache map;private String name;private String cache_name;private int autowarmCount;private State state;private CacheRegenerator regenerator;private String description="Eh LRU Cache";private static int cache_index = 0;private static Map<String, CacheManager> managerPool = null;private static Map<String, Integer> managerFlag = null;private static CacheManager managerTemplate = null;static{managerPool = new HashMap<String, CacheManager>();managerFlag = new HashMap<String, Integer>();managerTemplate = new CacheManager("/data/conf/ehcache.xml");}private Cache GetCache() {// use cache poolSet<String> set = managerFlag.keySet();Iterator<String> it = set.iterator();while(it.hasNext()) {String cacheName = it.next();if( managerFlag.get(cacheName) == 0 ) { // not usedmanager = managerPool.get(cacheName);System.out.println("EhCacheWrapper Cache Name(Pool): " + cacheName);managerFlag.put(cacheName, 1);cache_name = cacheName;return manager.getCache(cacheName);}}// add zhenjingString cacheName = name + cache_index;System.out.println("EhCacheWrapper Cache Name: " + cacheName);// create Cache from templateCache orig = managerTemplate.getCache(name);CacheConfiguration configTmp = orig.getCacheConfiguration();configTmp.setName(cacheName);Configuration managerConfiguration = new Configuration();managerConfiguration.setName(cacheName);manager = new CacheManager(managerConfiguration.cache(configTmp));// put to cache poolmanagerFlag.put(cacheName, 1);managerPool.put(cacheName, manager);// get cachecache_index++;cache_name = cacheName;return manager.getCache(cacheName);}public Object init(Map args, Object persistence, CacheRegenerator regenerator) {state=State.CREATED;this.regenerator = regenerator;name = (String)args.get("name");String str = (String)args.get("size");final int limit = str==null ? 1024 : Integer.parseInt(str);str = (String)args.get("initialSize");final int initialSize = Math.min(str==null ? 1024 : Integer.parseInt(str), limit);str = (String)args.get("autowarmCount");autowarmCount = str==null ? 0 : Integer.parseInt(str);// get cachemap = GetCache();CacheConfiguration config = map.getCacheConfiguration();description = "Eh LRU Cache(MaxBytesLocalOffHeap=" + config.getMaxBytesLocalOffHeap() + ", MaxBytesLocalHeap=" + config.getMaxBytesLocalHeap() + ", MaxEntriesLocalHeap=" + config.getMaxEntriesLocalHeap() + ")";if (persistence==null) {// must be the first time a cache of this type is being createdpersistence = new CumulativeStats();}stats = (CumulativeStats)persistence;return persistence;}public String name() {return name;}public int size() {synchronized(map) {return map.getSize();}}public Object put(Object key, Object value) {synchronized (map) {if (state == State.LIVE) {stats.inserts.incrementAndGet();}// increment local inserts regardless of state???// it does make it more consistent with the current size...inserts++;map.put(new Element(key,value));return null; // fake the previous value associated with key.}}public Object get(Object key) {synchronized (map) {Element val = map.get(key);if (state == State.LIVE) {// only increment lookups and hits if we are live.lookups++;stats.lookups.incrementAndGet();if (val!=null) {hits++;stats.hits.incrementAndGet();//System.out.println(name + " EH Cache HIT. key=" + key.toString());}}if( val == null) return null;return val.getObjectValue();}}public void clear() {synchronized(map) {map.removeAll();}}public void setState(State state) {this.state = state;}public State getState() {return state;}public void warm(SolrIndexSearcher searcher, SolrCache old) throws IOException {return;}public void close() {clear();// flag un-used managerFlag.put(cache_name, 0);System.out.println("EhCacheWrapper Cache Name(Reuse): " + cache_name);}SolrInfoMBeans methods //public String getName() {return EhCacheWrapper.class.getName();}public String getVersion() {return SolrCore.version;}public String getDescription() {return description;}public Category getCategory() {return Category.CACHE;}public String getSourceId() {return " NULL ";}public String getSource() {return " NULL ";}public URL[] getDocs() {return null;}// returns a ratio, not a percent.private static String calcHitRatio(long lookups, long hits) {if (lookups==0) return "0.00";if (lookups==hits) return "1.00";int hundredths = (int)(hits*100/lookups); // rounded downif (hundredths < 10) return "0.0" + hundredths;return "0." + hundredths;/*** code to produce a percent, if we want it...int ones = (int)(hits*100 / lookups);int tenths = (int)(hits*1000 / lookups) - ones*10;return Integer.toString(ones) + '.' + tenths;***/}public NamedList getStatistics() {NamedList lst = new SimpleOrderedMap();synchronized (map) {lst.add("lookups", lookups);lst.add("hits", hits);lst.add("hitratio", calcHitRatio(lookups,hits));lst.add("inserts", inserts);lst.add("evictions", evictions);lst.add("size", map.getSize());}lst.add("warmupTime", warmupTime);long clookups = stats.lookups.get();long chits = stats.hits.get();lst.add("cumulative_lookups", clookups);lst.add("cumulative_hits", chits);lst.add("cumulative_hitratio", calcHitRatio(clookups,chits));lst.add("cumulative_inserts", stats.inserts.get());lst.add("cumulative_evictions", stats.evictions.get());return lst;}public String toString() {return name + getStatistics().toString();} }

外部ehcache.xml配置：

<ehcache xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"xsi:noNamespaceSchemaLocation="http://ehcache.org/ehcache.xsd"updateCheck="false" monitoring="autodetect"dynamicConfig="true" name="config"><!--<cache name="filterCache"maxEntriesLocalHeap="1024"eternal="true"overflowToOffHeap="true"maxBytesLocalOffHeap="1g"></cache><cache name="fieldValueCache"maxEntriesLocalHeap="1024"eternal="true"overflowToOffHeap="true"maxBytesLocalOffHeap="1g"></cache>--><cache name="queryResultCache"maxEntriesLocalHeap="1"eternal="true"overflowToOffHeap="true"maxBytesLocalOffHeap="800m"></cache><!-- ehcache not support documentCache, encoding format error.<cache name="documentCache"maxEntriesLocalHeap="1024"eternal="true"overflowToOffHeap="true"maxBytesLocalOffHeap="1g"></cache>--></ehcache>

http://www.cnblogs.com/zhenjing/p/Ehcache_BigMemory.html

總結

以上是生活随笔為你收集整理的Ehcache BigMemory: 摆脱GC困扰(转)的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。