You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
yacy_search_server/source/de/anomic/server/serverMemory.java

224 lines
9.7 KiB

// serverMemory.java
// -------------------------------------------
// (C) 2005 by Michael Peter Christen; mc@yacy.net, Frankfurt a. M., Germany
// first published 22.09.2005 on http://yacy.net
//
// $LastChangedDate$
// $LastChangedRevision$
// $LastChangedBy$
//
// LICENSE
//
// This program is free software; you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation; either version 2 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License for more details.
//
// You should have received a copy of the GNU General Public License
// along with this program; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
package de.anomic.server;
import de.anomic.server.logging.serverLog;
* Complete number localization and provide a more reasonable interface to serverObjects: - put(key, value) methods are now used if a value added to the map should be kept as it is. Numbers are transformed (but not formatted) to an equivalent String representation. - putASIS(...) have been removed, now done with simple put(...) (see above). - puNum(...) can be used for number values which should be stored in a formatted way, either depending on the current locale setting for yacy (default) or in a "none" locale (see javadocs and setLocalize()). - putHTML(...) escapes special characters into corresponding HTML enities ('<' => '&lt;') which was done with put(...) before and so was called too often, becauses it is necessary only for very few cases. Additionally there is a "forXML" mode which only replaces < > & ". In short: Use put(...) for almost everything, use putXY(...) if you need some special transformation of the value. A few bugs have been fixed as well, and there should be a small performance improvement for complex pages with a lot of values. * added additional Sum/Avg rows to access tracker pages, see http://forum.yacy-websuche.de/viewtopic.php?f=5&t=456 * removed duplicate code (mostly related to the big changes above). TODO: - make sure, number formats work as expected _everywhere_, report overseen stuff http://forum.yacy-websuche.de/viewtopic.php?f=5&t=437 - probably a good idea to add special putDate() methods as they are used in many pages and create duplicated formatting code + maybe some centralized handling for memory value formatting. - further improve the speed of page creation for the WatchCrawler. git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@4178 6c8d7289-2bf4-0310-a012-ef5d649a1542
17 years ago
import de.anomic.tools.yFormatter;
public class serverMemory {
public static boolean vm14 = System.getProperty("java.vm.version").startsWith("1.4");
public static final long max = (vm14) ? computedMaxMemory() : Runtime.getRuntime().maxMemory() ; // patch for maxMemory bug in Java 1.4.2
private static final Runtime runtime = Runtime.getRuntime();
private static final serverLog log = new serverLog("MEMORY");
private static final long[] gcs = new long[5];
private static int gcs_pos = 0;
private static long lastGC;
public final synchronized static void gc(int last, String info) { // thq
long elapsed = System.currentTimeMillis() - lastGC;
if (elapsed > last) {
long free = free();
System.gc();
lastGC = System.currentTimeMillis();
if (log.isFine()) log.logInfo("[gc] before: " + bytesToString(free) + ", after: " + bytesToString(free()) + ", call: " + info);
} else if (log.isFine()) {
log.logFinest("[gc] no execute, last run: " + (elapsed / 1000) + " seconds ago, call: " + info);
}
}
/** @return the amount of freed bytes by a forced GC this method performes */
private static long runGC(final boolean count) {
final long memnow = available();
gc(1000, "serverMemory.runGC(...)");
final long freed = available() - memnow;
if (count) {
gcs[gcs_pos] = freed;
gcs_pos = (gcs_pos + 1) % gcs.length;
}
return freed;
}
/**
* This method calculates the average amount of bytes freed by the last GCs forced by this class
* @return the average amount of freed bytes of the last forced GCs or <code>0</code> if no
* GC has been run yet
*/
public static long getAverageGCFree() {
long x = 0;
int y = 0;
for (int i=0; i<gcs.length; i++)
if (gcs[i] != 0) {
x += gcs[i];
y++;
}
return (y == 0) ? 0 : x / y;
}
public static long free() {
// memory that is free without increasing of total memory taken from os
return runtime.freeMemory();
}
/**
* Tries to free a specified amount of bytes. If the currently available memory is enough, the
* method returns <code>true</code> without performing additional steps. Otherwise - if
* <code>gciffail</code> is set - a Full GC is run, if <code>gciffail</code> is set to
* <code>false</code> and not enough memory is available, this method returns <code>false</code>.
*
* @see serverMemory#request(long, boolean) for another implementation
* @param memory amount of bytes to be assured to be free
* @param gciffail if not enough memory is available, this parameter specifies whether to perform
* a Full GC to free enough RAM
* @return whether enough RAM is available
* @deprecated use {@link serverMemory#request(long, boolean)} instead to enable the collection of
* heuristics about explicit usage of Full GCs.
*/
public static boolean available(long memory, boolean gciffail) {
if (available() >= memory) return true;
if (!gciffail) return false;
gc(4000, "serverMemory.available(...)");
return (available() >= memory);
}
public static long available() {
// memory that is available including increasing total memory up to maximum
return max - runtime.totalMemory() + runtime.freeMemory();
}
/**
* <p>Tries to free a specified amount of bytes.</p>
* <p>
* If the currently available memory is enough, the method returns <code>true</code> without
* performing additional steps. If not, the behaviour depends on the parameter <code>force</code>.
* If <code>false</code>, a Full GC is only performed if former GCs indicated that a GC should
* provide enough free memory. If former GCs didn't but <code>force</code> is set to <code>true</code>
* a Full GC is performed nevertheless.
* </p>
* <p>
* Setting the <code>force</code> parameter to false doesn't necessarily mean, that no GC may be
* performed by this method, if the currently available memory doesn't suffice!
* </p>
* <p><em>Be careful with this method as GCs should always be the last measure to take</em></p>
*
* @param size the requested amount of free memory in bytes
* @param force specifies whether a GC should be run even in case former GCs didn't provide enough memory
* @return whether enough memory could be freed (or is free) or not
*/
public static boolean request(final long size, final boolean force) {
long avail = available();
if (avail >= size) return true;
if (log.isFine()) {
String t = new Throwable("Stack trace").getStackTrace()[1].toString();
log.logFine(t + " requested " + (size >> 10) + " KB, got " + (avail >> 10) + " KB");
}
final long avg = getAverageGCFree();
if (force || avg == 0 || avg + avail >= size) {
// this is only called if we expect that an allocation of <size> bytes would cause the jvm to call the GC anyway
final long freed = runGC(!force);
avail = available();
log.logInfo("performed " + ((force) ? "explicit" : "necessary") + " GC, freed " + (freed >> 10)
+ " KB (requested/available/average: " + (size >> 10) + " / "
+ (avail >> 10) + " / " + (avg >> 10) + " KB)");
return avail >= size;
} else {
log.logInfo("former GCs indicate to not be able to free enough memory (requested/available/average: "
+ (size >> 10) + " / " + (avail >> 10) + " / " + (avg >> 10) + " KB)");
return false;
}
}
public static long used() {
// memory that is currently bound in objects
return runtime.totalMemory() - runtime.freeMemory();
}
public static String bytesToString(long byteCount) {
try {
final StringBuffer byteString = new StringBuffer();
if (byteCount > 1073741824) {
* Complete number localization and provide a more reasonable interface to serverObjects: - put(key, value) methods are now used if a value added to the map should be kept as it is. Numbers are transformed (but not formatted) to an equivalent String representation. - putASIS(...) have been removed, now done with simple put(...) (see above). - puNum(...) can be used for number values which should be stored in a formatted way, either depending on the current locale setting for yacy (default) or in a "none" locale (see javadocs and setLocalize()). - putHTML(...) escapes special characters into corresponding HTML enities ('<' => '&lt;') which was done with put(...) before and so was called too often, becauses it is necessary only for very few cases. Additionally there is a "forXML" mode which only replaces < > & ". In short: Use put(...) for almost everything, use putXY(...) if you need some special transformation of the value. A few bugs have been fixed as well, and there should be a small performance improvement for complex pages with a lot of values. * added additional Sum/Avg rows to access tracker pages, see http://forum.yacy-websuche.de/viewtopic.php?f=5&t=456 * removed duplicate code (mostly related to the big changes above). TODO: - make sure, number formats work as expected _everywhere_, report overseen stuff http://forum.yacy-websuche.de/viewtopic.php?f=5&t=437 - probably a good idea to add special putDate() methods as they are used in many pages and create duplicated formatting code + maybe some centralized handling for memory value formatting. - further improve the speed of page creation for the WatchCrawler. git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@4178 6c8d7289-2bf4-0310-a012-ef5d649a1542
17 years ago
byteString.append(yFormatter.number((double)byteCount / (double)1073741824 ))
.append(" GB");
} else if (byteCount > 1048576) {
* Complete number localization and provide a more reasonable interface to serverObjects: - put(key, value) methods are now used if a value added to the map should be kept as it is. Numbers are transformed (but not formatted) to an equivalent String representation. - putASIS(...) have been removed, now done with simple put(...) (see above). - puNum(...) can be used for number values which should be stored in a formatted way, either depending on the current locale setting for yacy (default) or in a "none" locale (see javadocs and setLocalize()). - putHTML(...) escapes special characters into corresponding HTML enities ('<' => '&lt;') which was done with put(...) before and so was called too often, becauses it is necessary only for very few cases. Additionally there is a "forXML" mode which only replaces < > & ". In short: Use put(...) for almost everything, use putXY(...) if you need some special transformation of the value. A few bugs have been fixed as well, and there should be a small performance improvement for complex pages with a lot of values. * added additional Sum/Avg rows to access tracker pages, see http://forum.yacy-websuche.de/viewtopic.php?f=5&t=456 * removed duplicate code (mostly related to the big changes above). TODO: - make sure, number formats work as expected _everywhere_, report overseen stuff http://forum.yacy-websuche.de/viewtopic.php?f=5&t=437 - probably a good idea to add special putDate() methods as they are used in many pages and create duplicated formatting code + maybe some centralized handling for memory value formatting. - further improve the speed of page creation for the WatchCrawler. git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@4178 6c8d7289-2bf4-0310-a012-ef5d649a1542
17 years ago
byteString.append(yFormatter.number((double)byteCount / (double)1048576))
.append(" MB");
} else if (byteCount > 1024) {
* Complete number localization and provide a more reasonable interface to serverObjects: - put(key, value) methods are now used if a value added to the map should be kept as it is. Numbers are transformed (but not formatted) to an equivalent String representation. - putASIS(...) have been removed, now done with simple put(...) (see above). - puNum(...) can be used for number values which should be stored in a formatted way, either depending on the current locale setting for yacy (default) or in a "none" locale (see javadocs and setLocalize()). - putHTML(...) escapes special characters into corresponding HTML enities ('<' => '&lt;') which was done with put(...) before and so was called too often, becauses it is necessary only for very few cases. Additionally there is a "forXML" mode which only replaces < > & ". In short: Use put(...) for almost everything, use putXY(...) if you need some special transformation of the value. A few bugs have been fixed as well, and there should be a small performance improvement for complex pages with a lot of values. * added additional Sum/Avg rows to access tracker pages, see http://forum.yacy-websuche.de/viewtopic.php?f=5&t=456 * removed duplicate code (mostly related to the big changes above). TODO: - make sure, number formats work as expected _everywhere_, report overseen stuff http://forum.yacy-websuche.de/viewtopic.php?f=5&t=437 - probably a good idea to add special putDate() methods as they are used in many pages and create duplicated formatting code + maybe some centralized handling for memory value formatting. - further improve the speed of page creation for the WatchCrawler. git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@4178 6c8d7289-2bf4-0310-a012-ef5d649a1542
17 years ago
byteString.append(yFormatter.number((double)byteCount / (double)1024))
.append(" KB");
} else {
byteString.append(Long.toString(byteCount))
.append(" Bytes");
}
return byteString.toString();
} catch (Exception e) {
return "unknown";
}
}
private static int computedMaxMemory() {
// there is a bug in java 1.4.2 for maxMemory()
// see for bug description:
// http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4686462
// to get the correct maxMemory, we force a OutOfMemoryError here to measure the 'real' maxMemory()
int mb = 1024 * 1024;
byte[][] x = new byte[2048][];
for (int i = 0; i < x.length; i++) {
try {
x[i] = new byte[mb];
} catch (OutOfMemoryError e) {
x = null; // free memory
//System.out.println("* computed maxMemory = " + i + " mb");
return (int) Math.max(i * mb, Runtime.getRuntime().totalMemory());
}
}
return 2048 * mb;
}
public static void main(String[] args) {
// try this with a jvm 1.4.2 and with a jvm 1.5 and compare results
int mb = 1024 * 1024;
System.out.println("vm: " + System.getProperty("java.vm.version"));
System.out.println("computed max = " + (computedMaxMemory() / mb) + " mb");
int alloc = 10000;
Runtime rt = Runtime.getRuntime();
byte[][] x = new byte[100000][];
for (int i = 0; i < 100000; i++) {
x[i] = new byte[alloc];
if (i % 100 == 0) System.out.println("used = " + (i * alloc / mb) +
", total = " + (rt.totalMemory() / mb) +
", free = " + (rt.freeMemory() / mb) +
", max = " + (rt.maxMemory() / mb) +
", avail = " + ((rt.maxMemory() - rt.totalMemory() + rt.freeMemory()) / mb));
}
}
}