Traditional Clusters






Why Traditional Clusters?
If your goal is investment protection of your existing infrastracture you should consider Appro's rack-mounted line of servers. Appro's outstanding thermal management and innovative mechanical design allow efficient airflow for multiple processors in a high-density platform. The benefit of choosing traditional rack mounted servers is commodity based, high performance computing in an industry
standard form factor. Appro's server clusters
feature multiple CPU capability, ease of management,
large high speed SMP memory capacity and large
internal storage capacity. Appro's turn-key cluster
solutions offers high-speed interconnect options
including Myrinet, Infiniband, Dolphin and Quadrics.
Appro's servers provide enterprise class reliability
and scalability with outstanding processing foundation
to support applications such as Beowulf clusters,
financial modeling, compute farms, digital rendering,
seismic analysis, video streaming, web hosting,
complex custom application development, and others
requiring higher floating-point and memory bandwidth
performance.
For your turn-key cluster solutions, Appro two product series:
XtremeServers based on Next-Generation AMD Opteron Processors
with DDR2 and HyperServers based on the latest Dual-Core Intel
Xeon Processors. These servers are ideal for memory-intensive
and I/O-intensive applications and for customers who need a
more rich configuration in an innovative density-optimized server
with simplified management at an excellent performance.
Learn more about Appro's servers:

Appro
XtremeServers - AMD Opteron

Appro
HyperServers - Intel® Xeon®

Performance/Watt
of Traditional Clusters
Choosing an optimal power solution requires the consideration
of many factors, including the number of server enclosures needed,
the type of servers to be installed, and the type of available
power. Appro 's servers feature an innovative system design
for power efficiency, superior airflow for multiple processors
in a high-density platform as well as the ability to utilize
low power CPU's offering an excellent performance per watt ratio.
1U Clusters vs Blades
Blades servers typically consist of a motherboard,
a hard drive, one or two processors, memory, network
connections, and sometimes an optional back-up
drive that mirrors the primary drive. Blade servers
capitalize on the fact that all servers share
common components: they take the power supplies,
fans, network switches, and system ports of traditional
servers and relocate them to a back plane on the
blade-server chassis. While it is true that you
can fit more individual CPUs per cabinet using
blade servers, the power consumption drastically
increases per cabinet. Although some blade servers
do use lower power CPUs, the total power consumption
per cabinet is typically an order of magnitude
higher than what is consumed with high-density
1U Appro server clusters and Appro HyperBlade
Clusters.
Features and Benefits of Traditional Clusters
The Appro IU Cluster can address the needs of
most HPC applications. Its key features include
outstanding performance; reliability and superior
mechanical and thermal design derived from the
Appro IU servers:
Industry standard, commodity based server with
flexible 1U options
Commodity based server with flexible 1U, 2U and
3U configuration options
Rear door cabinet cooling option for improved
cooling in your data center
Large memory capacity and flexibility to support
a variety of applications
Redundant power supplies and fans reduces the
risk of failure
Low voltage, mid voltage and full power processors available
Cost effective, scalable performance for HPC and
enterprise applications
Up to 40 compute nodes per cabinet
8 Memory sockets per processor available
1U Servers with 2GB memory modules can go up to 32GB at DDR2
AMD Opteron and Intel Xeon processors
PCI-Express and PCI-X support
42U rack cabinet
Appro ServerDome Remote Management
High-Speed Interconnect Options available
Configuration options including, SAN, NAS based
storage
Choice of Linux or Windows operating systems
|