aries.5 (2010 09)

a
ARIES(5) ARIES(5)
(HP Integrity Systems Only)
NOTE: Stack sizes can be inherited among Integrity system native processes (32-bit and 64-bit) and
emulated PA-RISC processes (32-bit and 64-bit), subjected to the limitation on available stack space
under ARIES.
ARIES PERFORMANCE
For comparison purpose, the following machine configurations are assumed:
Integrity system: Model rx4640, Itanium2 4 CPU, clock speed 1500 MHz, 6 MB L3 cache, 8 GB
memory
HP9000 system: Model rp4440, PA8800 2 CPU dual core, clock speed 1000 MHz, 32 MB off chip
cache, 8 GB memory
The following table shows relative performance of PA-RISC applications under ARIES on an Integrity
system at 1500 MHz CPU clock speed.
Application Class Compared to Compared to
Integrity HP9000
system system
General (mix of I/O, memory, 60-70 % 80-110 %
system intensive ops)
Integer intensive 45-60 % 80-110 %
Floating point 20-30 % 40-60 %
intensive
Java based 40-70 % 50-80 %
applications
Multi threaded 50-60 % 70-80 %
applications
The performance numbers shown in above table are only indicative. The actual performance of your
PA-RISC application may vary depending on the application execution profile. You should benchmark
your application performance under ARIES on Integrity systems before actual deployment.
The class of applications that experience good performance under ARIES:
1. General applications, which have average mix of I/O intensive, memory intensive and system inten-
sive operations.
2. Integer intensive applications which spend significant time in algorithmic computations.
3. Any category of PA-RISC applications migrated from old PA-RISC systems like N-class and earlier.
4. Static image rendering graphics applications (no animations).
5. Certain data base and client applications.
The class of applications that experience poor performance under ARIES:
1. Floating point intensive applications like most scientific simulation, modeling applications. The per-
formance slowdown is due to architectural differences between PA-RISC and Itanium(R) processor
features and resources for floating point operations.
2. Java-based applications. Unless your application has JNI code, use Integrity native JVM. Depend-
ing on execution profile, performance of Java applications varies significantly under ARIES.
3. Multi-threaded applications that spend significant amount of time in thread synchronization opera-
tions. Thread suspension operation experiences slow performance under ARIES.
4. PA-RISC applications which use performance libraries such as
HP mlib are not good candidates to
run under ARIES. It is not possible for ARIES to emulate the detailed behavior of the application for
maximum emulated hardware resource utilization.
5.
OpenGL-based PA-RISC applications experience poor performance under ARIES. Such applications
would execute with better performance under ARIES if they can use OpenGL display lists
and send graphics data using GLX protocol mode.
6. PA-RISC applications which frequently use
volatile variables or are compiled with option +Ovo-
latile.
HP-UX 11i Version 3: September 2010 − 13 − Hewlett-Packard Company 13