aim9 URL: http://www.caldera.com/developers/community/contrib/aim.html

Developed by AIM Technologies for testing performance of UNIX hardware. Suites VII and IX of the AIM benchmarks were placed under the GPL by Caldera. At that time they annonced a planned Sourceforge project for support - this has not appeared. The test measures various operations and returns results in terms of operations per second. The suite performs the following tests:

Sample Output with explanation.

Interpretation Guide

Download STP version of the test

Also see: Linux Benchmark Suite Homepage


Bash-memory URL: http://www.osdl.org/sw_resources/ext3-tools.tar.gz URL: http://www.uow.edu.au/~andrewm/ext3-tools.tar.gz

bash-memory places lots of virtual memory pressure on a Linux system. It consists of 2 programs: usemem and bash-shared-mapping. Each of these programs tries to consume all of the available RAM and cause much swapping (if swapping is enabled).

bash-memory is not a benchmark suite, it is a workload. If successful, it will complete in a reasonable amount of time. If it takes more than 24 hours to complete or it does not complete (system hangs, seen mostly in early 2.4 days), the test fails, but usually there is little or no notification of this.

Sample Output


bonnie++ URL: http://www.textuality.com/bonnie/

Bonnie is a benchmark which measures the performance of Unix file system operations. Works on a file of specified size (default 100Mb) reports bytes/sec, bytes/CPU sec, % CPU usage (user and system).

Sample Output

Interpretation Guide


Contest URL: http://members.optusnet.com.au/ckolivas/contest/

Contest is designed to test system responsiveness by running kernel compilation under a number of different load conditions.

Sample Output

Interpretation Guide


dbench (long and quick) URL: ftp://samba.org/pub/tridge/dbench/

Runs the standard dbench program for a long period of time, or once for a sanity check (dbench-quick) dbench produces only the filesystem load. It does all the same IO calls that the smbd server in Samba would produce when confronted with a netbench run. It does no networking calls.

Sample Output Measured output: table of throughput (Mb/sec) as
the number of clients increases.


DBT1-1tier URL:http://sourceforge.net/projects/osdldbt/

Commercial Web/Database load. A single tier configuration of an ecommerce OLTP database test that is derrived from the TPC-W benchmark. It simulates users browsing and ordering books from an on-line retailer. Three-tier model (driver, web server, database) run with single system. Load comprised running 300 users against a 4GB database of 1000 items. Run on a 2 or 4 CPU machine. Database backend: SAPDB; web server is simulated. I/O for database: raw. Data captured: transactions per second, average response time, I/O, CPU utilization, memory usage.

Sample Output with explainations

Detailed Information


DOTS URL: http://sourceforge.net/projects/ltp

The Database Opensource Test Suite is a set of stress and stability test cases designed for measuring databse performance and reliability. It performs coverage testing for JDBC APIs and models real-world business logic. The test run covers seven Basic cases and two Advance cases, where each case is a combination of query, update, insert and delete

Sample Output Advanced Case 1

Detailed Information


Hackbench Taken from the Linux Kernel Mailing list

The hackbench test is a benchmark for measuring the performance, overhead, and scalability of the linux scheduler. It uses a client and server processes grouped into so-called chatrooms, where each client sends a message to each server in the group.

The present wrapped version of the benchmark executes it with various #'s of groups in order to measure the scalability performance of the scheduler.

Sample Output

20 4.622
40 7.9854
60 12.8744666666667
80 19.0577
100 26.7034

The first column is the number of chat rooms. The second column shows the total time for the test run.

Commentary


IOZone URL: http://www.iozone.org/

The IOZONE Filesystem Benchmark utility, specifically testing for a specified filesystem.

Commentary

Detailed Information (pdf format)

Sample Output For the ext2 filesystem


LMbench URL: http://www.bitmover.com/lmbench/

From Larry McVoy and the BitMover team. A suite of 'simple, portable benchmarks' Includes:

The lmbench-short version omits the memory tests. We do not do the multi-node network portion.

Sample Output

Detailed Information


Linux Test Project URL: http://ltp.sourceforge.net/

The Linux? Test Project is a joint project with SGI, IBM®, OSDL, and Bull® with a goal to deliver test suites to the open source community that validate the reliability, robustness, and stability of Linux. The Linux Test Project is a collection of tools for testing the Linux kernel and related features. Our goal is to improve the Linux kernel by bringing test automation to the kernel testing effort. Interested open source contributors are encouraged to join the project.

Sample Output List of failed tests

Detailed Information


Memtest URL: http://carpanta.dc.fi.udc.es/~quintela/memtest/

The memtest suite put significant memory pressure on the Linux VM subsystem and cause much swapping when using Linux versions 2.4.15 and 2.4.16.

memtest is not a benchmark suite, it is a workload. If successful, it will complete in a reasonable amount of time. If it takes more than 24 hours to complete or it does not complete (system hangs, seen mostly in early 2.4 days), the test fails, but usually there is little or no notification of this.

Sample Output


pgmeter URL: http://pgmeter.sourceforge.net/

PenguinoMeter (also known as pgmeter) is a file-I/O benchmark for Linux. It is designed to measure the rate at which data can be transferred to a file as opposed to the rate at which file-system operations (such as create file, find file, delete file, etc) can be performed.

Commentary

Sample Output

Detailed Information


Tiobench URL: http://tiobench.sourceforge.net/

tiobench is a fully-threaded I/O benchmark program. It is configured to run using 3 to 8 times the system RAM size for its target file size. tiobench can be run on any of these filesystems in OSDL STP: ext2, ext3, reiserfs, JFS, or XFS.

Sample Output


UNIXbench URL: http://www.tux.org/pub/tux/benchmarks/System/unixbench/

The BYTE UNIX benchmark test suite

Sample Output


VM_Regress URL: http://www.csn.ul.ie/~mel/projects/vmregress/

VM Regress is a regrssion, testing and benchmark tool for the Linux Kernel Virtual Memory (VM) system. The first significant problem with developing with the VM is that testing it is extremly tricky. Simulating scenarios reliably is difficult at best and impossible at worst. This leads to the tendancy to rely on stress testing to reveal flaws or to write quick custom tests that only one developer can avail of.

The second problem is that emperical performance data is virtually impossible to obtain and doing a direct comparison between two VM's relies on subjective data. This makes is impossible to prove one VM is better over another or that a code change or feature introduction is beneficial.

VM Regress is a tool that will ultimatly be able to perform reliable, reproducable testing on every aspect of the VM including providing empirical performance data. It relies on a combination of userspace tools and kernel modules to accuratly determine the current state of the kernel and to reliably reproduce tests. Testing with such accuracy from userspace is difficult and obtaining test results is similar to watching shadows on a wall.

Modules are divided up into 4 section. core modules provide infrastructure for the whole suite. sense reveal information about the current running kernel. test test specific code paths and scenarios. bench provide benchmarks for different subsystems.

VM Regress provides an interface via the proc system. A directory entry is created in /proc/vmregress and new entries are created there as modules are loaded. Generally a small help will be available by cating the entries but the manual or .c file should be consulted for extra information.

Sample Output