AIM Multiuser Benchmark

OVERVIEW
The AIM Multiuser Benchmark - Suite VII tests and measures the performance of
Open System multiuser computers. Multiuser computer environments typically have
the following general characteristics in common: 

- A large number of tasks are run concurrently 
- Disk storage increases dramatically as the number of users increase. 
- Complex numerically intense applications are performed infrequently 
- An important amount of time is spent sorting and searching through large
  amounts of data. 
- After data is used it is placed back on disk because it is a shared resource. 
- A large amount of time is spent in common runtime libraries. 

Multiuser systems are commonly used to support the following types of user
environments: 

- Multiuser/shared system environment performing office automation, word
  processing, spreadsheet, email, database, payroll, and data processing. 
- Compute server environment that uses an extremely large quantity of data,
  performs large quantities of floating point calculations, and large amounts
  of Interprocess Communications (IPC) for graphics. 
- Large database environment with a lot of disk I/O, data in memory, and IPC
  via shared memory.
- File server environment with a heavy concentration of integer compute file
  system operations. 


The AIM Multiuser Benchmark, also called the AIM Benchmark Suite VII or AIM7, 
is a job throughput benchmark widely used by UNIX computer system vendors.
The original code was from AIM Technology, Inc., who licensed it to others. 
Caldera International, Inc., bought the license and released the source code 
for Suites VII and IX under the GPL.

AIM7 is a C program that forks many processes which represent jobs or users. 
Each job is composed of as much as 53 assorted tests blended to create a 
workload that exercises a different aspect of the operating system such as 
disk-file operations, process creation, user virtual memory operations, 
pipe I/O, and compute-bound arithmetic loops. The test proportions are 
specified via a workfile used to define the workload.

A complete AIM7 benchmark run is comprised of a series of independent runs of 
the selected workload at different requested loads, specified in terms of a 
number of jobs. Each individual run executes until all of its jobs have 
completed the set of randomly ordered tests specified by the workfile. A number 
of metrics describing the results at that load point are reported including 
the rate at which the system under test was able to complete the work, or the
number of jobs completed per minute. The metric of greatest interest is peak 
system throughput, the throughput obtained at some requested load (in terms 
of a number of jobs per minute) that was greater than the throughput obtained 
for all other requested loads. I.E., a given system will have a peak number of 
tasks N at which the jobs per minute is maximized.  Either N, or the value of 
the jobs per minute at N, is considered the peak system throughput.

The number of requested jobs per load point defaults to increasing by one, 
however using the adaptive option the number of requested jobs can increase 
by much more than one.

The AIM suite provides several examples of these workloads including 
simulations of databases, file servers, and compute servers. As mentioned the 
workload can be adjusted by altering test weight or modifying the test mix in 
the workfile.

The default workloads for AIM7 are compute loads (CPU scalability), shared 
users (VM and file systems), database workload (mix weighted toward disks 
random I/O), and file server (mix weighted towards sequential and random disk 
I/O).

The workloads will continue to add user processes where each process runs a 
mix of operations. A metric for the number of jobs per minute (jobs/min) 
represents the throughput for the system under test (SUT). A balanced system 
should allow server memory, disks, and file systems to be added to the SUT 
until the number of processes exceeds the number of jobs/min. This metric is 
called the AIM7 "crossover-point" or when sustained throughput equals the 
jobs/min. Historically this was considered an excellent measure of performance 
because many times a system's expandability does not match the hardware level 
and its ability for the OS to scale. However, as system's performance has
increased (better JPM throughput) it can take many hours for a workload to
reach the "crossover-point". As such, we mainly focus on the JPM metric
and the scalability graphs of Users vs JPM.


LICENSE
This software is distributed under the GNU General Public License - see the
accompanying COPYING file for more details.

SUPPORTED OPERATING SYSTEMS
This software was designed to run properly on any POSIX-compliant operating
system. An ANSI C compiler is required.

DOCUMENTATION
The original AIM documentation has been restored from a printed copy. The
original document was scanned, OCR'ed, edited for contents and converted
manually into HTML format. The base document is "doc/main.html."

INSTALLATION
Please refer to the INSTALL file.

TRADEMARKS
"AIM Benchmark" and "Hot Iron Awards" are trademarks of Caldera
International, Inc. (Caldera) in the USA and other countries. Caldera allows
use of "AIM" and "AIM Benchmarks" when presenting performance results, but
places restrictions on this to promote comparability of results and avoid
confusion. Results presented using the terms "AIM" and "AIM Benchmark" must be
obtained with an official version of the benchmark, as designated by Caldera,
using one of the standard distributed workload mixes. No right or license is
granted hereunder to use the Caldera trademarks for modified benchmark versions
or for other purposes without the prior written permission of Caldera. Should
you need to contact Caldera about that topic, please e-mail
aimbench@sourceforge.net.

"UNIX" is a trademark of the The Open Group.