How to design workload model for stress testing?

  In this blog post, we will see how to design workload model for stress testing. Every person in this world has a breaking point. In short, there is a limit for everything. It applies for everything. Similarly, servers also has a breaking point, beyond a particular limit it will not behave as expected. It is much important to identify the breaking point before application goes live.

See the definition of Stress testing in my earlier post Types of Performance Testing. Stress Testing validates the stability of the system and it identifies the breaking point by overloading the concurrent users.

We will see what the appropriate approach for stress testing is. Following are the entry criteria for stress testing.

• Hardware configurations
• Network configurations
• Number of Concurrent Users derived from Load Testing i.e. peak load
• Identified critical scenarios

It is best to refer hardware and network configuration, as it will help to plan the stress testing to avoid any disaster.

Consider a web application which publishes your post across social medias; total number of concurrent during peak business hours (from 11:00 to 13:00) per day is 1500. Following are the list of transactions performed by the users.

1. Login
2. Click on Write Post link
3. Write Post (140 characters)
4. Submit Post
5. Logout

In order to identify the breaking point of the system, first you load the server with 1500 concurrent users i.e. at A as shown in below figure.

Load the server with 1500 concurrent users gradually. Once it reached, increase the concurrent users to 1600 and observe the script execution. Monitor the graphs between Response Time vs. Number of Virtual Users (VUsers), Throughput vs. Number of VUsers, and Errors vs. Number of VUsers. These graphs alerts the breaking point.

If everything goes fine for 1600 VUsers, increase it to 1700 VUsers, 1800 VUsers and so on till you face any abnormal behavior of the server. At some point, i.e. in our scenario at point B say at 2000, there are abnormal behavior such as Error 404, Error 500 errors occurred during the execution.

It is ideal to stop the execution either gradually or immediately. Because further execution give no valid result for analysis. In this case, breaking point of the server is 2000, i.e. 2000 VUsers can login concurrently and access the website. But beyond 2000, server will not be able to handle those many users.

To summarize, first is to gather entry criteria and start loading the server and incrementing it, till it breaks the system.

How to use performance center ?

 In this blog post, we will see about how to use Performance Center for Load/Stress Testing. HP Performance Center is a web based application which enables performance test engineers to execute load/stress testing.

The main advantage of Performance Center is any number of users can access, plan, execute and analyze the results.

The knowledge of LoadRunner components such as VuGen, Controller and Analysis is preferable. Now we will see steps involved to conduct load / stress testing using Performance Center. If you noticed, I did not mention any version number of Performance Center. If I mentioned version here, this blog post will be outdated in near future. Here we are going to see core steps to conduct load / stress testing using Performance Center.

Steps involved in Performance Center to conduct Load / Stress Testing

a. Creating scripts using VuGen

b. Uploading scripts to Performance Center

c. Host Machines

d. Creating Scenarios

e. Booking Slots

f. Conducting Load / Stress Testing

g. Analyze Results

a. Creating scripts using VuGen

From Performance Center, VuGen tool can downloaded for creating scripts for the identified critical scenarios. Recording, Playback, enhancing scripts and unit test will yield base lined VuGen script for testing.

b. Uploading scripts to Performance Center

Base lined scripts should be uploaded to Performance Center. Scripts should be uploaded in *.zip format this can be achieved thru File menu in VuGen.

c. Host machines

Before conducting load / stress tests, we will see about host machines which acts as Controller which manages the load / stress test. Load Generator which actually generate traffic by executing Virtual Users (VUsers). Controller + Load Generator which acts as Controller and Load Generator. Above details can be seen at Timeslot page.

d. Creating Scenarios

By clicking on New Load Test in Load Test page, one can create scenarios for Load / Stress testing. Under Design Group, test engineers can specify the number of load generators which can set either by number or by percentage.

Scheduler is used to specify the number of virtual users for the load / stress tests. It allows test engineers to specify the duration, ramp-up, ramp-down, steady state span and run mode. It displays an interactive schedule graph for the test engineers.

e. Booking slots

As I mentioned in first paragraph, Performance Center is a web based application. Test Engineers across geographical location will be able to conduct load / stress testing. It is much important to book the slots, so that no conflict will raise during execution.

Timeslot displays the availability of load generators, controllers and VUsers. Test engineers need to specify the time slot and the required hosts to conduct his/her load / stress test.

f. Conducting Load / Stress Testing

To execute load test, go to Scheduler and click on Start button to execute load / stress test or click on Run button against the load test. Performance Center scans for any errors. If everything is green, we can see the number of VUsers running, passed/failed transactions details, errors etc.

g. Analyzing the results

Once the test gets completed, we can collate and analyze the result. Results will be available in *.html, results.zip and rawresults.zip for analysis. Bottlenecks can be identified using correlating one or more graphs. The results can be downloaded and exported to Analysis tool for detailed analysis.

In next post, we will see about the best practices and the general errors that occurs during load / stress test.

Correlation issue - conversion from hex to string

void convert_hexparam(const char* paramName)
{
	char* src=0;
	char* psrc=0;
	char* dst = 0;
	char* pdst = 0;

	size_t srclen = 0;

	char c;
	char buf[5];

	char enclosedParamName[256];

	sprintf(enclosedParamName, "{%s}", paramName);
	src=lr_eval_string(enclosedParamName);

	srclen = strlen(src);

	dst = (char*)malloc(srclen);
	memset(dst, 0, srclen);

	for(psrc=src, pdst = dst; *psrc != '\0'; pdst++)
	{
		if(*psrc== '\\' && *(psrc + 1) == 'x')
		{
			memset(buf, 0, 5);
			buf[0] = '0';
			buf[1] = 'x';
			buf[2] = *(psrc + 2);
			buf[3] = *(psrc + 3);
			c = (char)strtol(buf, 0, 0 );

			*pdst  = c;
			psrc += 4;
		}
		else
		{
			*pdst = *psrc;
			++psrc;
		}	
	}
	*(pdst+1) = '\0';

	lr_save_string(dst, paramName);

	free(dst);
}

Action()
{
	lr_save_string("tjIAKJprJ\\x2FxoK2GJL\\x2BZMhgUm9Wat4j16dw\\x2B7o4d9Lak\\x3D", "hex_data");

	lr_output_message(lr_eval_string("{hex_data}"));

	convert_hexparam("hex_data");

	lr_output_message(lr_eval_string("{hex_data}"));

	return 0;
}

Random Name Generator

keep in global.sh:


random_alpha(char* param_name, int length) 
{ 
char buff[32] = ""; 
int r,i; 
char c; 
srand((unsigned int)time(0)); //Seed number for rand() 

for (i = 0; i < 7; i++) 
{ 
// A-Z = 65-90 = rand() % 25 + 65 

r = rand() % 25 + 65; 
c = (char)r; 

buff[i] = c; 
printf("%c", c); 

} 

lr_save_string(buff, param_name); 
} 


random_alpha("P_FirstName", 5);

Capture SAMLResponse dynamic value in loadrunner script

#ifndef _GLOBALS_H
#define _GLOBALS_H

//--------------------------------------------------------------------
// Include Files
#include "lrun.h"
#include "web_api.h"
#include "lrw_custom_body.h"


        void solution (char *strSource, char *strFrom, char *strTo, char *strParam)
       
        {
       
        char strTemp[29999];
       
        char *s = (char*)strSource;
       
        int i=0;
       
        int iLenFrom = strlen (strFrom);
       
        int iLenTo = strlen (strTo );
       
       
        while (*s)
       
        {
       
        if (strncmp (s, strFrom, iLenFrom) == 0)
       
        {
       
        strncpy (&(strTemp[i]), strTo, iLenTo);
       
        i += iLenTo;
       
        s += iLenFrom;
       
        }
       
        else
       
        {
       
        strTemp[i++] = *s;
       
        s++;
       
        }
       
        }
       
        strTemp[i] = 0;
       
       
        lr_save_string (strTemp, strParam);
       
        }

#endif // _GLOBALS_H


Variable declaration in action :

Action ()

{

char *strSource,*strSource1;
   

char *sOut;

    web_set_max_html_param_len("698956");

     web_reg_save_param("samlreq_1","LB=<input name=\"SAMLResponse\" type=\"hidden\" value=\"","RB=\"/>",LAST);

    web_url("sessionCookieRedirect",
        "URL=https://faxxx/login/sessionCookieRedirect?checkAccountSetupComplete=true&token={token}&redirectUrl=https%3A%2F%2m%2Fhome%2_eagentsiteminderperf_1%2F0oaz0twsmBgXpBqpw2p6%2Falnz0yo44UQb1s93c2p6",
        "TargetFrame=",
        "Resource=0",
        "RecContentType=text/html",
        "Referer=https://eagenttl.farmersinsurance.com/login/forms/login.fcc?fromURI=https%3A%2F%2%2Fhome%2Ffinctest_eagentsiteminderperf_1%2F0oaz0twsmBgXpBqpw2p6%2Falnz0yo44UQb1s93c2p6%3FTYPE%3D33554433%26REALMOID%3D06-3e8f9418-cdf1-0093-0000-408000004080%26GUID%3D%26SMAUTHREASON%3D0%26METHOD%3DGET%26SMAGENTNAME%3D-SM-o3pMQdyFKmzLzrop3mcsQBRiC%252fLMGjgA9LYV%252bVhWi%252buwxN1IOW0b2LkK3slO9EtZmT%252b9kLM%252fOjI0KfNunPgfmQKxH0RuraDz%26TARGET%3D-SM-https%253a%252f%252feagenttl%252e%252ecom%252f",
        "Snapshot=t9.inf",
        "Mode=HTML",
        LAST);

    web_url("saml",
        "URL=https://xxxx/app/xxx_eagentsiteminderperf_1/exkz0twslaWHTtbr72p6/sso/saml",
        "TargetFrame=",
        "Resource=0",
        "RecContentType=text/html",
        "Referer=",
        "Snapshot=t10.inf",
        "Mode=HTML",
        LAST);

   
   
    strSource=lr_eval_string("{samlreq_1}");
       
        //lr_save_string (strSource,"pNewValue0"); 
       
        solution (strSource,"&#x2b;","+","samlreq_1");
       
        strSource=lr_eval_string("{samlreq_1}");
       
        //lr_save_string (strSource,"postValue1"); 
       
        solution (strSource,"&#x3d;","=","C_Samlresponse_Converted");

    web_submit_data("saml2assertionconsumer",
        "Action=https://xxxxx/affwebservices/public/saml2assertionconsumer",
        "Method=POST",
        "TargetFrame=",
        "RecContentType=text/html",
        "Referer=https://xxxx/app/finctest_eagentsiteminderperf_1/exkz0twslaWHTtbr72p6/sso/saml",
        "Snapshot=t11.inf",
        "Mode=HTML",
        ITEMDATA,
        "Name=SAMLResponse", "Value={C_Samlresponse_Converted}", ENDITEM,
        "Name=RelayState", "Value=", ENDITEM,
        LAST);
   

}

Oracle NCA Protocol Example Scripts

In the following example, the user selected an item from a list (nca_list_activate_item), pressed a button (nca_button_press), retrieved a list value (nca_lov_retrieve_items), and performed a click in an edit field (nca_edit_click). The logical names of the objects are the parameters of these functions.

nca_lov_select_item("Responsibilities","General Ledger, Vision Operations");
nca_list_activate_item("FNDSCSGN.NAVIGATOR.LIST.0","+ Journals"); 
nca_list_activate_item("FNDSCSGN.NAVIGATOR.LIST.0"," Enter"); 
nca_button_press("GLXJEENT.TOOLBAR.LIST.0"); 
nca_lov_find_value("Batches",""); 
nca_lov_retrieve_items("Batches",1,9); 
nca_lov_select_item("Batches","AR 1020 Receivables 2537: A 1020"); 
nca_edit_click("GLXJEENT.FOLDER_QF.BATCH_NAME.0");

In certain tests, such as those performed on Oracle Configurator applications, information returned by one function is required throughout the session. VuGen automatically saves the dynamic information to a parameter by inserting a web_reg_save_param function into the script.

In the following example, the connection information is saved to a parameter called NCAJServSessionID. The right boundary is \r. The actual right boundary may differ between systems.

web_reg_save_param ("NCAJServSessionId", "LB=\r\n\r\n", "RB=\r",
       LAST);
web_url("f60servlet",         
         "URL=http://usscifforms05.sfb.na/servlet/f60servlet\?config=mult",
         LAST);

Oracle NCA Protocol Example Scripts

In the following example, the user selected an item from a list (nca_list_activate_item), pressed a button (nca_button_press), retrieved a list value (nca_lov_retrieve_items), and performed a click in an edit field (nca_edit_click). The logical names of the objects are the parameters of these functions.

... 
nca_lov_select_item("Responsibilities","General Ledger, Vision Operations");
nca_list_activate_item("FNDSCSGN.NAVIGATOR.LIST.0","+ Journals");
nca_list_activate_item("FNDSCSGN.NAVIGATOR.LIST.0","       Enter");
nca_button_press("GLXJEENT.TOOLBAR.LIST.0");
nca_lov_find_value("Batches","");
nca_lov_retrieve_items("Batches",1,9);
nca_lov_select_item("Batches","AR 1020 Receivables 2537: A 1020");
nca_edit_click("GLXJEENT.FOLDER_QF.BATCH_NAME.0");
...

In certain tests, such as those performed on Oracle Configurator applications, information returned by one function is required throughout the session. VuGen automatically saves the dynamic information to a parameter, by inserting a web_reg_save_param function into the script. In the following example, the connection information is saved to a parameter called NCAJServSessionID.

web_reg_save_param ("NCAJServSessionId", "LB=\r\n\r\n", "RB=\r",
       LAST);
web_url("f60servlet",         
         "URL=http://usscifforms05.sfb.na/servlet/f60servlet\?config
         =mult",         LAST);

In the above example, the right boundary is \r. The actual right boundary may differ between systems.

Loadrunner And Oracle Part1

Introduction

This is part 1 of an article which discusses some of the mechanisms for performance testing Oracle Applications / Oracle Forms using LoadRunner. This is far from an in-depth article but will, we hope, help performance testers get started with testing Oracle Applications and act as a reference resource. The article as a whole covers two key areas that seem to present the most issues, Object Names and Connect issues, as well as some other issues that we at Testing Performance have come across when delivering our Performance Testing Services. Part 1 covers Object Names issues.

Object Names

One of the first issues you will come across when recording an Oracle Application is that Oracle objects are recorded by a number rather than a name. These numbers are dynamic and cannot be relied upon when playing back a script, i.e. the object cannot be found by LoadRunner as its “number” has changed. Here is an example of objects recorded by their object number (this is part of a login script that was recorded using Oracle Applications 11i protocol against an Oracle Forms application): … nca_edit_set(“31”, ”USERNAME”); nca_obj_type(“31”, ‘\t’, 0); nca_edit_set(“32”, (”PASSWORD”)); nca_obj_type(“32”, ‘\t’, 0); nca_edit_set(“33”, ”TRAINING”); nca_button_press(“34”); … This script refers to four objects 31, 32, 33 and 34. If these objects do not change their number between recording and execution then the script should run OK. However these numbers are subject to change for example if a new field is added on the form between object 33 and object 34 then the nca_button_press(“34”) would fail as would now probably known as 35. To get round this issue Oracle offers the “RECORD=NAMES” method which associates a name to each object rather than a number. The name is very much more likely to remain static. Here is the same example as above except the objects were recorded by their object name (this is part of a login script that was recorded using Oracle Applications 11i protocol against an Oracle Forms application): … nca_edit_set(“LOGON_INFO_USER_NAME”, ”USERNAME”); nca_obj_type(“LOGON_INFO_USER_NAME”, ‘\t’, 0); nca_edit_set(“LOGON_INFO_USER_PASSWORD”,(”PASSWORD”)); nca_obj_type(“LOGON_INFO_USER_PASSWORD”,(””, ‘\t’, 0); nca_edit_set(“LOGON_INFO_USER_DATABASE”,(””,”TRAINING”); nca_button_press(“LOGON_INFO_CONNECT_BUTTON”); … By default, in all the engagements that I have been involved in at least, Record=Names is not enabled. So how do you enable Record=Names? Here are four methods for achieving this, three of which require a configuration change and one which is very simple. I have found that finding the correct method is very much a “suck it and see” methodology. Before trying the following just make sure that your installation records numbers instead of names by recording a simple logon script.

Method 1: Append to URL

This is very much the simplest method and should be tried first. Below is the Start Recording dialog presented when recording using the Oracle Applications 11i protocol. The URL in this case is: http://apps.acme.co.uk:7778/forms/frmservlet?config=2001 Try appending record=names to the end of the URL, for example: http://apps.acme.co.uk:7778/forms/frmservlet?config=2001&record=names Note if record=names is the first parameter to be added to the url then you will need a question mark before it eg: http://apps.acme.co.uk:7778/forms/frmservlet?record=names If however there are one or more parameters already in the url then you will need the ampersand prefix eg: http://apps.acme.co.uk:7778/forms/frmservlet?config=2001&record=names Now try and record a simple login script, if the object numbers are replaced by object names then you are in business. You should also notice that the record=names flag is appended to the connect_server statement. Note: This method worked with the following configuration: Oracle Forms version: 1012002 LR Protocol: Oracle Applications 11i LR Version: 9.5 Jinitiator version: 1.3.1.28 (not overly important)

Method 2: Set record=names in startup file

This method requires a change to the HTML startup file on the forms server. Tracking which file to change can be a bit tricky if someone has renamed the appropriate file, which they are quite entitled to do. The file you are looking for is the one that is called when the forms applet starts. This file holds configuration details for the forms applet. By default the name of the file is basejini.htm. The file contains a number of configuration entries.  The line you are interested in is something like this: <PARAM name="serverArgs ... ... ... fndnam=APPS"> [param name="”serverArgs" fndnam="APPS”" /] Note there may be a number of entries which replace the ellipses (…) above. Append record=names as the last argument i.e.: [param name="”serverArgs" fndnam="APPS" record="names”" /] <PARAM name="serverArgs ... ... ... fndnam=APPS record=names"> Note: there is a space between the fndnam=APPS argument and record=names. Now try and record a simple login script, if the object numbers are replaced by object names then you are in business. Note: This method worked with the following configuration: Oracle Forms version: 60824 LR Protocol: Oracle Applications 11i LR Version: 9.5 Jinitiator version: 1.1.8.16 (not overly important)

Method 3: Set value in formsweb file and HTML start-up file

I have had to use this method less often than the previous two methods. This method should be used if the HTML start-up file references the Forms CGI configuration file.  This method requires two files to be changed. The Forms CGI configuration file in question is formsweb.cfg (it may have been renamed but probably not). You will need to add a new entry in this file under the USER PARAMETERS in the section: “Values for Forms applet parameters”. Add the following as the last entry in this section: xrecord=names Your file should look something like this: … connectMode=socket serverHost=serv1.acme.co.uk xrecord=names ;4) Parameters fo Jinitiator … In the HTML start-up file basejini.htm (or whatever its name has been changed to) there are a number of configuration entries.  The lines you are interested in go something like this: <PARAM name="serverArgs ... ... ... fndnam=APPS"> [param name="”serverArgs" fndnam="APPS”" /] Note there may be a number of entries which replace the ellipses (…) above. Append record=%xrecord% as the last argument i.e.: [param name="”serverArgs" fndnam="APPS" record="%xrecord%”" /] <PARAM name="serverArgs ... ... ... fndnam=APPS record=%xrecord%"> Note 1: there is a space between the fndnam=APPS argument and record=%xrecord%. Note 2: there may be more than one entry in the HTML start-up file that needs to be changed

Method 4: Oracle Applications record=names flag

If you can access the Oracle Applications System administrator then you can change the profile of the users via the Profile System Values panel and appending ?play=&record=names to the ICX:Forms Launcher. At this time I have no screen shots for this process but please feel free to contact Testing Performance ( info@testingperformance.co.uk) should you need assistance and we will do our best to provide a more indepth procedure.   Part 2 coming soon! Part 2 of this article covers Connect issues and other issues that you may encounter when doing performance testing with Oracle NCA– look out for it! Remember if you need help with any Performance Testing Services please feel free to contact us.

Multi-Protocol (Oracle NCA/Web) Tips & Tricks

Pre-Patch 4 Architecture

Pre-Patch 4 Architecture issues for Internet Deployment of Forms

Forms Listener Servlet – patch 4

Recording a Multi-Protocol NCA/Web Script

Replaying a Multi-Protocol NCA/Web Script

Correlating a Multi-Protocol NCA/Web Script

1. edit_get_text

2. java_get_value

Tips and Tricks

Q.    What version of LoadRunner supports Multi-Protocol NCA/Web script?

Q.    How to enable Web replay’s Run-Time Settings.

Q.    When to use Multi-Protocol Oracle NCA/Web.

Q.    How to enable Multi-Protocol in LoadRunner 7.5 and LoadRunner7.51.

Q.    How to replay a Multi-Protocol NCA/Web script on Unix.

Q.    Error: "The formsLauncher failed to connect to the existing forms session properly…” on recording

Tuning Garbage Collection Outline

This document is a summary or outline of Sun's document: Tuning Garbage collection with the 1.4.2 Hotspot JVM located here: http://java.sun.com/docs/hotspot/gc1.4.2/

1.0 Introduction

• For many applications garbage collection performance is not significant
• Default collector should be first choice

2.0 Generations

• Most straightforward GC will just iterate over every object in the heap and determine if any other objects reference it.
• This gets really slow as the number of objects in the heap increase
• GC's therefor make assumptions about how your application runs.
• Most common assumption is that an object is most likely to die shortly after it was created: called infant mortality
• This assumes that an object that has been around for a while, will likely stay around for a while.
• GC organizes objects into generations (young, tenured, and perm) This is important!

2.1 Performance Considerations

• Ways to measure GC Performance
• Throughput - % of time not spent in GC over a long period of time.
• Pauses - app unresponsive because of GC
• Footprint - overall memory a process takes to execute
• Promptness - time between object death, and time when memory becomes available
• There is no one right way to size generations, make the call based on your applications usage.

2.2 Measurement

• Throughput and footprint are best measured using metrics particular to the application.
• Command line argument -verbose:gc output
  [GC 325407K->83000K(776768K), 0.2300771 secs]
• GC - Indicates that it was a minor collection (young generation). If it had said Full GC then that indicates that it was a major collection (tenured generation).
• 325407K - The combined size of live objects before garbage collection.
• 83000K - The combined size of live objects after garbage collection.
• (776768K) - the total available space, not counting the space in the permanent generation, which is the total heap minus one of the survivor spaces.
• 0.2300771 secs - time it took for garbage collection to occur.
• You can get more detailed output using -XX:+PrintGCDetails and -XX:+PrintGCTimeStamps

3 Sizing the Generations

• The -Xmx value determines the size of the heap to reserve at JVM initialization.
• The -Xms value is the space in memory that is committed to the VM at init. The JVM can grow to the size of -Xmx.
• The difference between -Xmx and -Xms is virtual memory (virtually committed)

3.1 Total Heap

• Total available memory is the most important factor affecting GC performance
• By default the JVM grows or shrinks the heap at each GC to keep the ratio of free space to live objects at each collection within a specified range.
• -XX:MinHeapFreeRatio - when the percentage of free space in a generation falls below this value the generation will be expanded to meet this percentage. Default is 40
• -XX:MaxHeapFreeRatio - when the percentage of free space in a generation exceeded this value the generation will shrink to meet this value. Default is 70
• For server applications
• Unless you have problems with pauses grant as much memory as possible to the JVM
• Set -Xms and -Xmx close to each other or equal for a faster startup (removes constant resizing of JVM). But if you make a poor choice the JVM can't compensate for it.
• Increase memory sa you increase # of processors because memory allocation can be parallelized.

3.2 The Young Generation

• The bigger the young generation the less minor GC's, but this implies a smaller tenured generation which increases the frequency of major collections.
• You need to look at your application and determine how long your objects live for to tune this.
• -XX:NewRatio=3 - the young generation will occupy 1/4 the overall heap
• -XX:NewSize - Size of the young generation at JVM init. Calculated automatically if you specify -XX:NewRatio
• -XX:MaxNewSize - The largest size the young generation can grow to (unlimited if this value is not specified at command line)

3.2.1 Young Generation Guarantee

• The -XX:SurvivorRatio option can be used to tune the number of survivor spaces.
• Not often important for performance
• -XX:SurvivorRatio=6 - each survivor space will be 1/8 the young generation
• If survivor spaces are too small copying collection overflows directly into the tenured generation.
• Survivor spaces too large uselessly empty
• -XX:+PrintTenuringDistribution - shows the threshold chosen by JVM to keep survivors half full, and the ages of objects in the new generation.
• Server Applications
• First decide the total amount of memory you can afford to give the virtual machine. Then graph your own performance metric against young generation sizes to find the best setting.
• Unless you find problems with excessive major collection or pause times, grant plenty of memory to the young generation.
• Increasing the young generation becomes counterproductive at half the total heap or less (whenever the young generation guarantee cannot be met).
• Be sure to increase the young generation as you increase the number of processors, since allocation can be parallelized.

4 Types of Collectors

• Everything to this point talks about the default garbage collector, there are other GC's you can use
• Throughput Collector - Uses a parallel version of the young generation collector
• -XX:+UseParallelGC
• Tenured collector is the same as in default
• Concurrent Low Pause Collector
• Collects tenured collection concurrently with the execution of the app.
• The app is paused for short periods during collection
• -XX:+UseConcMarkSweepGC
• To enable a parallel young generation GC with the concurrent GC add -XX:+UseParNewGC to the startup. Don't add -XX:+UseParallelGC with this option.
• Incremental Low Pause Collector
• Sometimes called Train Collector
• Collects a portion of the tenured generation at each minor collection.
• Tries to minimize large pause of major collections
• Slower than the default collector when considering overall throughput
• Good for client apps (my observation)
• -Xincgc
• Don't mix these options, JVM may not behave as expected.

4.1 When to use Throughput Collector

• Large number of processors
• Reduces serial execution time of app, by using multiple threads for GC
• App with lots of threads allocating objects should use this with a large young generation
• Server Applications (my observation)

4.2 The Throughput collector

• By default the throughput collector uses the number of CPU's as its value for number of GC threads.
• On a computer with one CPU it will not perform as well as the default collector
• Overhead from parallel execution (synchronization costs)
• With 2 CPU's the throughput collector performs as well as the default garbage collector.
• With more then 2 CPU's you can expect to see a reduction in minor GC pause times
• You can control the number of threads with -XX:ParallelGCThreads=n
• Fragmentation can occur
• Reduce GC threads
• Increase Tenured Generation size

4.2.1 Adaptive Sizing

• Keeps stats about GC times, allocation rates, and free space then sizes young and tenured generation to best fit the app.
• J2SE 1.4.1 and later
• -XX:+UseAdaptiveSizePolicy (on by default)

4.2.2 Aggressive Heap

• Attempts to make maximum use of physical memory for the heap
• Inspects computer resources (memory, num processors) and sets params optimal for long running memory allocation intensive jobs.
• Must have at least 256MB of RAM
• For lots of CPU's and RAM, but 1.4.1+ has shown improvements on 4-Way machines.
• -XX:+AggressiveHeap

4.3 When to use the Concurrent Low Pause Collector

• Apps that benefit from shorter GC pauses, and can share resources with GC during execution.
• Apps with large sets of long living data (tenured generation)
• Two or more processors
• Interactive apps with modest tenured generation size, and one CPU

4.4 The Concurrent Low Pause Collector

• Uses a separate GC thread to do parts of the major collection concurrently with the app threads.
• Pauses App threads in the beginning of a collection and toward the middle (longer pause in middle)
• The rest of the GC is in a single thread that runs at the same time as the app

4.4.1 Overhead of Concurrency

• Doesn't provide much of an advantage on single processor machines.
• Fragmentation can occur.
• Two processor machine eliminates pauses due to the GC thread.
• The more CPU's the advantages of concurrent collector increase.

4.4.2 Young Generation Guarantee

• There has to be enough contiguous space available in the tenured generation for all objects in the eden and one survivor space.
• A larger heap is needed compared to the default collector.
• Add the size of the young generation to the tenured generation.

4.4.3 Full Collections

• If the concurrent collector is unable to finish collecting the tenured generation before the tenured generation fills up, the application is paused and the collection is completed.
• When this happens you should make some adjustments to your GC params

4.4.4 Floating Garbage

• Floating Garbage - Objects that die while the GC is running (after they have been checked).
• Increase the tenured generation by 20% to reduce floating garbage.

4.4.5 Pauses

• First Pause - marks live objects - initial marking
• Second Pause - remarking phase - checks objects that were missed during the concurrent marking phase due to the concurrent execution of the app threads.

4.4.6 Concurrent Phases

• Concurrent Marking phase occurs between initial mark and remarking phase.
• Concurrent sweeping phase collects dead objects after the remarking phase.

4.4.7 Measurements with the Concurrent Collector

• Use -verbose:gc with -XX:+PrintGCDetails
• vCMS-initial-mark shows GC stats for the initial marking phase
• CMS-concurrent-mark - shows GC stats for concurrent marking phase.
• CMS-concurrent-sweep - shows stats for concurrent sweeping phase
• CMS-concurrent-preclean - stats for determining work that can be done concurrently
• CMS-remark - stats for the remarking phase.
• CMS-concurrent-reset - concurrent stuff is done, ready for next collection.

4.4.8 Parallel Minor Collection Options with Concurrent Collector

• -XX:+UseParNewGC - for multiprocessor machines, enables multi threaded young generation collection.
• -XX:+CMSParallelRemarkEnabled - reduce remark pauses

4.5 When to use the Incremental Low Pause Collector

• Use when you can afford to tradeoff longer and more frequent young generation GC pauses for shorter tenured generation pauses
• You have a large tenured generation
• Single Processor

4.6 The Incremental Low Pause Collector

• Minor collections same as default collector.
• Don't use try to use parallel GC with this collector
• Incrementally Collects parts of the tenured generation at each young collection.
• Tries to avoid long major collections by doing small chunks each minor collection.
• Can cause fragmentation of the heap. Sometimes need to increase tenured generation size compared to the default.
• There is some overhead required to maintain the position of the incremental collector. Less overhead than is required by the default collector.
• First try the default collector, and adjust heap sizing. If major pauses are too long try incremental.
• If the incremental collector can't collect the tenured generation fast enough you will run out of memory, try reducing the young generation.
• If young generation collections do not free any space, could be because of fragmentation. Increase tenured generation size.

4.6.1 Measurements with the Incremental Collector

• -verbose:gc and -XX:+PrintGCDetails
• Look for the Train: to see the stats for the incremental collection.

5 Other Considerations

• The permanent generation may be a factor on apps that dynamically generate and load many classes (JSP, CFM application servers)
• You may need to increase the MaxPermSize, eg: -XX:MaxPermSize=128m
• Apps that rely on finalization (finalize method, or finally clauses) will cause lag in garbage collection. This is a bad idea, use only for errorious situations.
• Explicit garbage collection calls (System.gc()) force a major collection. You can measure the effectiveness of these calls by disabling them with -XX:+DisableExplicitGC
• RMI garbage collection intervals can be controlled with
• -Dsun.rmi.dgc.client.gcInteraval=3600000
• -Dsun.rmi.dgc.server.gcInterval=3600000
• On Solaris 8+ you can enable libthreads, lightweight thread processes, these may increase thread performance.
• To enable add /usr/lib/lwp to LD_LIBRARY_PATH
• Soft References cleared less aggressively in server.
• -XX:SoftRefLRUPolicyMSPerMB=10000
• Default value is 1000, or one second per MB

6 Conclusion

• GC can be bottleneck in your app.

More Information

• FAQ: http://java.sun.com/docs/hotspot/gc1.4.2/faq.html
• Concurrent collector research paper: http://research.sun.com/techrep/2000/abstract-88.html
• HotSpot VM command line options: http://java.sun.com/docs/hotspot/VMOptions.html