SPECvirt sc2013 User'sGuide

SPECvirt_sc2013 User‘sGuide

Version 1.00 - February17, 2015

• 1.0 Requirements
  • 1.1 Client requirements
  • 1.2 VM requirements
    • 1.2.1 Storage requirements
    • 1.2.2 VM creation and network configuration
    • 1.2.3 VM software requirements
• 2.0 Setting up the test environment
  • 2.1 SPECpoll setup on all VMs
  • 2.2 Infraserver VM workload setup
  • 2.3 Webserver VM workload setup
  • 2.4 Mailserver VM workload setup
    • 2.4.1 Creating user accounts on the mailserver
  • 2.5 Database server VM workload setup
  • 2.6 Application server VM workload setup
  • 2.7 Batchserver VM workload setup
• 3.0 Setting up the client
  • 3.1 Install Java JDK
  • 3.2 Set up client hostnames
  • 3.3 VM and client operating system tuning
  • 3.4 Setup SPECvirt harness
• Appendix A - How to Build BeSIM
  • A.1 UNIX/LINUX
  • A.2 Windows
• Appendix B - Installing BeSim as a FastCGI for the Apache HTTP Server
  • B.1 FastCGI source compilation
  • B.2 Apache FastCGI Web server module: mod_fastcgi
  • B.3 Testing the BeSim FastCGI

1.0Requirements

This guide assumesthat you are using the SPECvirt_sc2013 version and that you have familiaritywith virtualization concepts and implementations. You require experience withthe installation, configuration, management, and tuning of your selectedhypervisor platform. You must know how to use your virtualization platform tocreate, administer, and modify virtual machines and allocate system resourcesincluding:

• Set up and add physical and internal networks
• Add virtual disks to VMs
• Use operating system and application software packages using tools such as yum and rpm to install, update, and configure:
  • J2EE application server software and JDBC libraries (for example GlassFish)
  • DBMS (for example PostgreSQL)
  • IMAP mail server (for example Dovecot)
  • web server (for example Apache, PHP, and fastcgi)
• Understand and run shell scripts
• Edit files with a text editor such as vi

1.1 Clientrequirements

A client machineshould be able to drive a complete tile‘s workload if it has the followingminimum characteristics:

• A 2010 or later processor with any core combination resulting in at least eight threads
• At least 8 GB of memory
• Network to accommodate slightly over 1 GbE bandwidth
• 24 GB of storage

You can use morethan one physical client, virtual client, or a combination to drive a tile.

Note: Due to network latency, we recommend that you create a separate,dedicated webserver client. Instructions for this are in Appendix B of the SPECvirt Client Harness User‘s Guide.

1.2 VMrequirements

When creating VMs,use your preferred virtualization software to create a virtual machine that hasenough virtual CPUs, virtual memory, virtualized network interfaces, andvirtual disk space to run your operating system and workload softwaresuccessfully. 

Each tile requires at least 110 GB storage per tile + 32 GB per database serverfor every four tiles. When configuring LUNs, consider using several physicaldisks in a RAID 5 or RAID 10 configuration. You do not need to co-locate atile‘s VMs on one LUN; VMs from different tiles can be located on a single LUN.

1.2.1 Storagerequirements

Each workload requires the followingstorage at a minimum:

1.2.2 VM creationand network configuration

Once your virtualmachine is created, install your preferred operating system in the newlycreated virtual machine. Configure a network interface to be connected to anexternal network over which the clients and VMs communicate. We recommendconfiguring another virtual network interface on the appserver VMs andwebserver VMs to be connected to an internal network that the VM can use tocommunicate with the dbserver and infraserver VMs respectively. Since thecommunications are intra-SUT, this internal network does not have to beassociated with a physical network port. For example, on the VM:

cat/etc/sysconfig/network-scripts/ifcfg-eth0
#public IP to/from clients
NAME=eth0
ONBOOT=yes
BOOTPROTO=static
NETMASK=255.255.0.0
IPADDR=172.168.122.64

cat /etc/sysconfig/network-scripts/ifcfg-eth1
#private intra-SUT IP communicates over Virtual NIC to other VMs only
NAME=eth1
ONBOOT=yes
BOOTPROTO=static
NETMASK=255.255.255.0
IPADDR=10.0.0.64 

1.2.3 VM softwarerequirements

All software usedmust be currently supported per the SPECvirt_sc2013 Run and Reporting Rules.

• For mailserver: IMAP server software that is IMAP4 rev 1 compliant
• For application server: J2EE compliant application server software (version 5 or later)
• For database server: DBMS server software that meets ACID (Atomicity, Consistency, Isolation, and Durability) requirements
• For web server and infraserver: Web server software that can support PHP and HTTP 1.1 (such as IIS, Apache, Zeus, Sun One)
• For infraserver and batch server: C compiler for compiling the BeSim on the infraserver and the CPU workload on the batch server
• For webserver: A supported version of PHP 5.x

On all VMs exceptfor the application server and batch server, install your preferred Java 6 orgreater Compatible JRE software. On the application server and batch serverVMs, install Java 6 or greater Compatible JDK software. A JRE is needed by theSPECpoll code that runs on all the VMs. SPECjAppServer2004 and SPEC CPU2006require a JDK. The Java JRE and JDK environments that are distributed with manyLinux distributions may not be sufficient.

To ensure you are using the correct version, issue:

java -version 

2.0 Setting upthe test environment

The SPECvirtbenchmark consists of four workloads:

• A web workload based on SPECweb2005 Support (webserver and infraserver)
• An IMAP workload based on SPECmail (mailserver)
• A Java application server workload based on SPECjAppServer2004 (appserver and dbserver)
• A batch workload based on SPEC CPU2006 "train" (batchserver)

While these fourworkloads are based on standalone SPEC benchmarks, the actual workloads used inSPECvirt are highly modified.  Therefore, the setup of these workloads inthe SPECvirt test environment differ somewhat from setting up the workloads intheir standalone form.  

Note: If you have a mix of Windows and Linux/Unix operating systems in yourSPECvirt configuration, it is recommended that you run a program such asdos2unix on all files modified under Windows and then copied toLinux/Unix.  This prevents file formatting errors which are difficult totrack down.

2.1 SPECpoll setupon all VMs

The pollme.jarlistener code of SPECpoll needs to be running on all VMs during the test andrequires Java 6 or greater Compatible JRE software. The Java runtimeenvironment that comes with many Linux distributions may not be sufficient.

Configure the operating system running on all workload VMs to start thelistener automatically on boot (for example, in /etc/rc.local), or have theclient start the listener as a part of the test run. The command to start onthis VM looks something like this: 

java -jar/opt/SPECpoll/pollme.jar -p 8001 

For more detailed information on SPECpoll and pollme.jar, see the SPECvirt Client Harness User‘s Guide.

2.2 Infraserver VMworkload setup

The infraserver VMcontains files that need to be accessed by the webserver VM. The infraserver VMalso acts as the back end simulator (BeSim) for the SPECweb2005 workload.

These are the steps for setting up the infraserver VM:

1.    Copy the SPECweb2005 directory from theharness kit onto a directory on the infraserver virtual machine.  (/optfor example)

2.    Setup a remote filesystem.

3.    Create a filesystem with 32 GB of space.

4.    Configure the filesystem to be accessibleby the webserver VM (for example, exporting the filesystem using NFS).

5.    To set up the BeSim functionality:

1.    Configure the system with Web serversoftware installed on port 81 or another port (avoid port 80 to avoid confusionwith the primary web server).

2.    Build your BeSim code as instructed in Appendix A. Select a BeSim implementation compatiblewith your Web server software (FastCGI*, ISAPI, or NSAPI), edit the appropriateMakeIncl.* file to reflect your compile and linker options, and select one ormore Defines. Then you can build the BeSim code for your selected API. (Ifusing FastCGI, copy the FastCGI executable from the Redistributed Open Sourcecode directory into the BeSim directory.)

3.    Install this code to your BeSim Webserver.

4.    Start your BeSim Web server.

5.    Use the test_besim_support scripts (included in the "besim" subdirectory of theinstallation) to test whether you‘re getting valid BeSim responses. Both Perlscripts and Linux bash scripts are provided. The Perl scripts require a Perlinterpreter as well as the following Perl modules: URI, HTML-Tagset,HTML-Parser, and libwww-perl. The bash scripts require cURL to be in the path.

Note: For moreFastCGI instructions (particularly for the Apache HTTP Server), see Appendix B. To download the FastCGI software:

wgethttp://www.fastcgi.com/dist/fcgi-2.4.0.tar.gz

If you have additional questions about setting up your BeSim server you maywant to refer to the SPECweb2005 User‘s Guide. Note that the SPECvirt harnessonly uses a modified version of the Support workload from SPECweb2005, so theEcommerce and Banking workloads do not need to be setup.

2.3 Webserver VMworkload setup

The webserver VMruns web server software that hosts a SPECweb2005-based Support workload usingSecure Sockets Layer (SSL). 

Make sure you perform the steps for the infraserver VM for your tile in theprevious section and power it on before setting up webserver. The webserver VMmust have the support/downloads directory remotely accessible from theinfraserver VM. This remote access point also needs to be present during webworkload file generation.

1.    Copy the SPECweb2005 directory from theharness kit onto a directory on the web server virtual machine.  (/optassumed for the purposes of the examples)

2.    Set localization on the system to US.

3.    Install and configure your favorite webserver software with PHP and SSL support. Note, the web server software shouldbe configured for HTTPS (port 443) and HTTP (port 80), which is used forinitialization.

4.    Set your current working directory to thewafgen directory:

cd/opt/SPECweb2005/wafgen

5.    Make the document root support/downloadsdirectory a fileserver mount from the infraserver VM. (You may want to use a VMinternal network for this mount if available.)

1.    Go to the SPECweb2005/wafgen directory andedit the support wafgen files to set the TILEINDEX value for the tile you arecreating.

2.    Edit /opt/SPECweb2005/wafgen/unix/support_downloads_props.rc and set the parameter TILEINDEX equal to the tile number minus 1 (forexample, set TILEINDEX=1 for tile 2).

3.    Edit /opt/SPECweb2005/wafgen/unix/support_image_props.rc and set the parameter TILEINDEX equal to the tile number minus 1 (forexample, set TILEINDEX=1 for tile 2)

4.    Build the image files:

./Wafgenunix/support_image_props.rc 

5.    Build the support files:

./Wafgen unix/support_downloads_props.rc 

Note that thiscommand could take over an hour to run.

6.  Set upPHP.

1.    Copy the contents of the /opt/SPECweb2005/Scripts/PHP/* directory to your document root:

cp -ar/opt/SPECweb2005/Scripts/PHP/* /var/www/html/ 

2.    Verify the "support" and"Smarty-2.6.26" subdirectories are under the document root. Ensurethe web server has read and write access to these directories andsubdirectories. From the document root:

chmod -R a+rw support/ Smarty-2.6.26/

An alternative isto use chown to assign ownership to the Web server user/group.

The stock PHP distribution ships with two ini files: php.ini-dist andphp.ini-recommended. While php.ini-recommended is optimized for performance andsecurity, for initial troubleshooting, php.ini-dist may be a better choice.This configures PHP to display errors in the browser. If PHP comes with your OSdistribution, you may only have one php.ini, in which case you should considerchanging "display_errors" and "display_startup_errors" to "On".  Youmay also want to set "error_reporting = E_ALL" to display all errors and warnings.

Since the datacreated for the webserver VM workload is specific to a particular tile, thedata files cannot be copied and used on the webserver VMs of different tiles.Each webserver must build its own image and support files.

If you have additional questions about setting up your web server, see theSPECweb2005 User‘s Guide. Note that the SPECvirt harness only uses a modifiedversion of the Support workload from SPECweb2005, so the Ecommerce and Bankingworkloads do not need to be setup.

2.4 Mailserver VMworkload setup

1.    Copy the SPECimap directory from theharness kit onto a directory on the mailserver virtual machine.  (/optassumed for the purposes of the examples)

2.    Install your preferred IMAP serversoftware.

3.    Configure a storage location in your IMAPserver that can hold 14 GB of data.

4.    Set up 500 users in your IMAP serversoftware for the benchmark. See below.

5.    Load the user accounts with data using theSPECvirt_sc2013 IMAP loader before you run the benchmark. The load may beinitiated from either a client machine or from the mailserver VM itself.

1.    Modify the IMAP config file(/opt/SPECimap/IMAP_config.rc)  for the load. Replace the CLIENTS variablewith the hostname of this mailserver VM followed by port 1200:

CLIENTS ="mailserver1:1200" 

A high thread count (such as 100) is recommended to expedite the loadingprocess in the next step. This can be set by changing the THREADS_PER_CLIENTline from 1 to a higher value; 100 has worked for some, but 20 may be a safervalue.

2.    Use the run_load_gen and run_init .sh or.bat files to load the mailstore: 

cd /opt/SPECimap
./run_load_gen.sh (in one virtualterminal or command shell)
./run_init.sh (in a secondvirtual terminal or command shell)

Note: This step populates the IMAP server with messages and may take more than20 minutes to complete.

Once the load isfinished, you need to kill the specimapclient Java process that is leftrunning.

Once a mailstore is created and loaded, it can be copied and used as themailstore on other tiles. We recommend backing up the mailstore once it has hadthe workload run against it and use this "warmed up" version of themailstore for the restores. This "warm up" period may need to be afew hours depending on your IMAP server.

2.4.1 Creatinguser accounts on the mailserver

You need to create500 user accounts on your mailserver for the benchmark and they must be of theform <USERNAME_PREFIX><USER_ID>, for example test1 and test500. TheUSERNAME_PREFIX is defined in the configuration file SPECimap/IMAP_config.rcand defaults to test. This parameter can be changed if needed to create usersin your IMAP server. Each user has a unique user ID which is a number whoselower and upper bounds are defined in the SPECimap/IMAP_config.rc configurationfile.

Note:  Some IMAP server software may also require you to createcorresponding operating system or domain user accounts.

2.5 Databaseserver VM workload setup

The dbserver VMacts as the database backend for the SPECjAppServer2004 workload. SPECvirt_sc2013 has modified the workload to access across 25 times moredatabase data than the usual SPECjAppServer2004 workload at a given injectionrate.  This means that the average SPECvirt_sc2013 driver injection rateof 1000 accesses a database sized for an injection rate of 500.

Note: The database server VM requires at least two vCPUs per tile.

1.    Copy the SPECjAppServer2004 from theharness kit onto a directory on the database server virtual machine. (/opt assumed for the purposes of the examples)

2.    Install your preferred database softwareand configure. The configuration for your database server software needs to betuned to handle the peak injection rate (IR) of 45.

3.    Configure a storage location in yourdatabase software that can hold 20 GB of data.

4.    Load your database for aSPECjAppServer2004 injection rate (IR) of 500. If desired, you can also runthese steps to load the database from the client rather than from the dbserverVM.

5.    Modify the/opt/SPECjAppServer2004/config/run.properties file on the client so thevariable txRate=500.

6.    Load the database using scripts specificto your database.  You can find example schemas and database populationscripts in the SPECjAppServer2004/schema directory. 

Note: To load the database successfully, consider temporarily editing thedatabase load script (loaddb.sh) to increase dbserver memory and vCPUs, Javaheap maximum size, and thread count.

7.    After the load finishes, ensure the valueof txRate in  /opt/SPECjAppServer2004/config/run.properties is at theruntime injection rate of txRate=100. Backup the database once it has beenbuilt.

The databasebackup can be copied and used on other tiles. A backup of the database isrequired because a database restore needs to be done before each benchmark run.

2.6 Applicationserver VM workload setup

Make sure you havedone the setup steps for the dbserver VM and that the dbserver VM for your tileis powered on before setting up the corresponding appserver VM. The appserverVM accesses the database contained on the dbserver VM.

1.    Copy the SPECjAppServer2004 directoriesfrom the harness kit onto a directory on the appserver server virtualmachine.  (/opt assumed for the purposes of  the examples)

2.    Install your preferred J2EE applicationserver software.  Note: For SPECvirt_sc2013, the Supplier Emulator(emulator.ear) MUST reside on the appserver VM.

3.    Setting up a SPECjAppServer2004environment is challenging. We strongly recommend using an application stackfrom a previously published SPECjAppServer2004 result. The SPECjAppServer2004Full Disclosure Archive associated with a result contains useful files andinformation that would help set up application and database server software.

Refer to theSPECjAppServer2004 User Guide to setup your J2EE application server for theSPECjAppServer2004 workload.  The database portion of the SPECjAppServertest is set up on the dbserver VM. 

The configuration for your J2EE application server needs to be tuned to handlea peak SPECjAppServer2004 injection rate (IR) of 200 (2,600 active webconnections).

2.7 Batchserver VMworkload setup

1.    Copy the SPECbatch directory from theharness kit onto a directory on the batch server virtual machine (/opt assumedfor the purposes of the examples).

2.    Make sure your preferred Java 6 or greaterCompatible JDK software is installed. The Java runtime environment that comeswith many Linux distributions may not be sufficient.

3.    Unpack the cpu2006-virt.tgz workloadarchive into the /opt/SPECbatch directory. After the extraction, you should seea cpu2006-virt directory.

4.    From the cpu2006-virt directory, run install.sh (install.bat for Windows). Accept the default options.

5.    Build the 401.bzip2 "base"executable. See the SPEC CPU2006 documentation for details and requirements for base optimizations.

6.    Customize your batch run script. Thesample run script virt_sc2013_run.sh is included in the kit and providesguidance in what is needed for such a script.

7.    Configure the VM‘s operating system tostart the SPECbatch listener automatically on boot (for example, in /etc/rc.local),or have the client start the listener as a part of the test run. The command tostart on this VM looks something like this: 

java -jar/opt/SPECbatch /batch.jar -p 1901

Note: You need acompiler present and configured on the VM. There are several exampleconfiguration files for different OS/compiler platforms available in theharness, and it is recommended that one of these be used.

3.0 Setting upthe client

3.1 Install JavaJDK

Install a Java 6or greater JDK on each client.

NOTE: The Java runtime environment that comes with many Linux distributions maynot work with the SPECvirt harness. It is recommended that you install anotherversion of Java that meets the 1.6 version minimum and ensure you are using thecorrect version by issuing:

java -version 

3.2 Set up clienthostnames

Entries need to be set up in the client‘shosts file (or equivalent) to associate the hostnames the SPECvirt harness usesto the appropriate IP addresses.

Web workload
Harness hostname defaults used:
infraserver - the server that acts as the Backend Simulator (BeSim) for the webworkload.
webserver - the server where the HTTP application is running

If you need to use a different hostname than the default then modify theWEB_SERVER parameter in the SPECweb2005/Test.config file accordingly once youhave installed the SPECvirt harness.

Note: We recommenddedicating a separate client to run only the webserver workload. See Appendix Bin the SPECvirt Client Harness User‘s Guide for details.

IMAP workload 
Harness hostname default used:
mailserver - the server where the IMAP application is running

If you need to use a different hostname than the default then modify theIMAP_SERVER parameter in the SPECimap/IMAP_config.rc file accordingly once youhave installed the SPECvirt harness. 

Appserver workload
Harness hostname defaults used:
specdelivery, appserver - the server where the application server is running
specemulator - the server where the emulator app is running (for SPECvirt this IP isthe same as specdelivery)
specdb, dbserver - the server where the database is running

If you need to use different hostnames than the defaults then refer to theSPECjAppServer2004 User‘s Guide document. 

Batch server workload 

Harness hostname default used:
batchserver - the server where the batch server is running

If you need to use a different hostname than the default then modify theBATCH_SERVER parameter in the SPECbatch/Test.config file accordingly once youhave installed the SPECvirt harness.

As an example, your hosts file entries should look something like this:

192.168.1.11infraserver 
192.168.1.12 webserver 
192.168.1.13 mailserver 
192.168.1.14 dbserver specdb 
192.168.1.15 appserver specdelivery specemulator 
192.168.1.16 batchserver 

3.3 VM and clientoperating system tuning

Running a SPECvirttile from a client probably requires that certain operating system tuning bedone in the following areas:

User Limits
If the operating system has any limits placed on the login session being usedto run the SPECvirt loads then those limits may need to be increased tosuccessfully complete a full SPECvirt benchmark run. For example, the followingmay need to be increased:

• Number of open files allowed by the user (ulimit -n)
• Number of processes (ulimit -u)

Networking
Operating system default network tunings may need to be increased to facilitatethe web workload. For example, the following may need to be increased:

• Transmit buffers
• Receive buffers

See submissions at SPECvirt_sc2013Results page for more detailedVM and client tuning.

3.4 Setup SPECvirtharness

See the SPECvirt Client Harness User‘s Guide for instructions on installing, configuring, and running the SPECvirtharness.

Appendix A -How to Build BeSIM

Instructions forbuilding BeSim on the infraserver follow.

A.1 UNIX/LINUX

To get name of theoperating system (for example, HP-UX):

uname -s 

In the Makefiles directory use one of the MakeIncl.<type>.<OS>files as a template for your  MakeIncl.<type>.<your_OS> andupdate to match your environment.
 
To build the FastCGI version of BeSim:  

make fcgi 
 
To build the NSAPI version of BeSim:   

make nsapi 
 
To build the Zeus ISAPI version of BeSim: 

make zisapi 
 
Optionally you can specify other make targets by adding TARGET=‘list oftargets‘. For example:
 
make zisapiTARGET=‘clean‘
make nsapi TARGET=‘clean all install‘  

The install target uses the path specified by DEST to install the BeSimexecutable or library, so it should be set to the appropriate place on yourBeSim webserver directory tree (i.e. /www/fast-cgi ).
 
If you plan to use FastCGI you must build and install the FastCGI developmentkit which provides the fast-cgi library and include files before buildingBeSim. If using Apache for the BeSim webserver, include mod_fastcgi, see http://www.fastcgi.com.

Note: Please send any new or updated  MakeIncl.* files you want includedin a future release to web2005(at)spec.org.

A.2 Windows

Project files areincluded as an example: BeSim.def, BeSim.dsp, BeSim.dsw

Appendix B -Installing BeSim as a FastCGI for the Apache HTTP Server

There are twopieces necessary to get BeSim for FastCGI working properly under Apache:

1.    Compiling the FastCGI source code and BeSim

2.    Compiling/configuring mod_fastcgi, anApache module to invoke FastCGI for specific directories

B.1 FastCGI sourcecompilation

To download theFastCGI software:

wgethttp://www.fastcgi.com/dist/fcgi-2.4.0.tar.gz

Compile the FastCGI source code.  The source tree is used on a BeSim(or full) installation of SPECweb2005. To compile it:

cd<path_to_SPECweb2005>/Besim/fcgi-2.4.0/ 
./configure --libdir=/usr/lib64 
make 
make install 

Note: --libdir=/usr/lib64 was added above due to the default FastCGI Makefileinstalling libraries to /usr/local, which is not a default library path onLinux and could cause this error upon execution of the FastCGI:

besim_fcgi.fcgi:error while loading shared libraries: 
libfcgi.so.0: cannot open shared object file: No such file or directory 

Create a directory for the BeSim FastCGI within your web server directoryhierarchy (but preferably outside the document root).  Here are somerecommended locations for different setups:

SuSE: /srv/www/fcgi-bin 
Red Hat: /var/www/fcgi-bin 
Apache 2.x compiled from source: /usr/local/apache2/fcgi-bin 

Compile BeSim as a FastCGI (replace /var/www/fcgi-bin/ with the directorycreated in the previous step):

cd<SPECweb2005>/Besim/make fcgi TARGET=‘clean all install‘DEST=/var/www/fcgi-bin/ 

This creates besim_fcgi.fcgi and install it into DEST.  If not, check themake output for errors.

B.2 Apache FastCGIWeb server module: mod_fastcgi

These instructionsshow how to compile the FastCGI web server module for Apache (mod_fastcgi). Forother web servers, please refer to their documentation to see whether FastCGIsupport is available.

Note for SuSE: SUSE Linux (8.1 and above) ships with a precompiled mod_fastcgifor Apache 2.x.  To check if it is installed, issue:

rpm -qapache2-mod_fastcgi 

If it‘s not installed, install this package from the CDs (see your OSdocumentation for more instructions).  If this package is installed, youcan skip the rest of this section.

Note for Red Hat: If using the httpd package that ships with the distribution,the httpd-devel RPM must also be installed as it contains the necessary utilities andheaders to compile Apache modules. To check, issue:

rpm -q httpd-devel 

Download the latest mod_fastcgi tarball from http://www.fastcgi.com/dist/mod_fastcgi-2.4.6.tar.gz. Follow thedirections in the INSTALL document (INSTALL.AP2 for Apache 2.x).

Note: As INSTALL.AP2 states, unless you are compiling from source, you need tomodify top_dir in your make command. For example, Red Hat users should use thefollowing ‘make install‘ command:

maketop_dir=/usr/httpd install 

You should now have a mod_fastcgi.so in your Apache modules directory:

Red Hat: 

ls -l /usr/lib/httpd/mod_fastcgi.so 
-rwxr-xr-x 1 root root 161408 Sep 16 13:47/usr/lib/httpd/modules/mod_fastcgi.so 

Compiled fromsource: 

ls -l /usr/local/apache2/modules/mod_fastcgi.so 
-rwxr-xr-x 1 root root 156434 Sep 21 10:03/usr/local/apache2/modules/mod_fastcgi.so 

Create a directory for FastCGI to store Unix socket files, and change thepermissions so the web server can have full access to it: 

Red Hat: 

mkdir -p/etc/httpd/fastcgi 
chmod 777 /etc/httpd/fastcgi 

Apache 2.x from source:

mkdir -p /usr/local/apache2/fastcgi 
chmod 777 /usr/local/apache2/fastcgi 
  
Add support for handling FastCGIs by making the following changes to httpd.conf(for Red Hat, this is in /etc/httpd/conf/):

1.    Add this line in the "Dynamic SharedObject (DSO) Support" section:

LoadModule fastcgi_module modules/mod_fastcgi.so 

2.      Add the following lines just afterthe ScriptAlias /cgi-bin/ line:

ScriptAlias /fcgi-bin/ "/var/www/fcgi-bin/" # <fcgi-bin directoryfrom Section 1, Step 2>
FastCgiIpcDir /etc/httpd/fastcgi # <FastCGI directory from previous step>

Note: Replace /var/www and /etc/httpd with /usr/local/apache2 if compiling from source. 

3.    Add the following section, preferablyafter the <Directory "<CGIDIR>/cgi-bin"> section:

<Directory"<CGIDIR>/fcgi-bin"> # <fcgi-bin directory from Section1, Step 2> 
   AllowOverride None 
   Options +ExecCGI -Includes 
   SetHandler fastcgi-script 
   Order allow,deny 
   Allow from all 
</Directory> 
AddHandler fastcgi-scriptfcgi

B.3 Testing theBeSim FastCGI

Start (or restart) your web server.

At the console, monitor your Apache error log to view any errors or warningsduring FCGI execution, i.e.

 tail -f<APACHE_ROOT>/logs/error_log 

You should initially see some FastCGI initialization messages such as:

[Tues Dec 1810:36:39 2012] [notice] FastCGI: process manager initialized (pid 25700) 

Test the BeSim FastCGI, using the scripts in the subdirectory of theSPECweb2005 installation. See Section 2.2 for examples.

SPEC and the namesSPECvirt, SPECvirt_sc, and SPECvirt_sc2013 are trademarks of the StandardPerformance Evaluation Corporation. Additional product and service namesmentioned herein may be the trademarks of their respective owners.

Copyright © 2010-2013 Standard Performance Evaluation Corporation (SPEC). Allrights reserved.