Xtreme-Cool Supercomputer
™
Xtreme-Cool Supercomputer
™
Overview
The Cray Xtreme-Cool supercomputer is based upon a liquid-cooling architecture
that directly cools the compute processor and memory. This results in superior
energy-efficiency over air-cooled systems by using warm water liquid-cooling heat
exchangers instead of chillers.
Less Energy, Lower TCO, Faster ROI
Requiring fewer or no air conditioning units, the liquid-cooled, Cray Xtreme-Cool
supercomputer can reduce typical energy consumption by 50% while dropping
power usage effectiveness (PUE) as low as 1.1. It offers a wide range of high dense
processor cores per rack. The Cray Xtreme-Cool system offers high performance
and three times more energy efficiency per rack compared to traditional air-cooled
designs.
Scalable and Optimized Architecture
The Cray Xtreme-Cool supercomputer can be configured in a Fat Tree or 3D Torus
architecture, optimizing it for superior application performance. The system is
composed of two processors per node supporting approximately 80 nodes per
rack in a standard 42U rack based on Intel® Xeon® processor E5 Family. It also
supports hybrid processing offering 40 nodes based on Intel® Xeon® processors
paired with Intel® Xeon Phi™ coprocessors or GPU accelerator nodes.
Ideal Environments
The Cray Xtreme-Cool design is perfect for medium-to-large high performance
computing (HPC) deployments supporting a variety of workload configurations.
It is an ideal solution for end users who require the best price/performance, scalability, reliability, manageability, and liquid-cooling energy efficient architecture
that reduces data center power and cooling operational costs for a lower total
cost of ownership (TCO).
Liquid-Cooled Energy-Efficient Architecture
•Reduces operating costs by using liquid-cooled heat exchangers
instead of chillers
•Uses low pressure isolated secondary loop to cool critical server
components
•Lowers energy costs, data center PUE of 1.1 or lower
•Directly cools the compute processor and memory
•Offers up to 80% heat capture in the warm water
•Produces high return water temperatures for possible heat reuse
•Offers leak detection and prevention features
•Integrated with remote monitoring and reporting system
•Fields more power dense systems in the data center
•Creates a better work environment through quieter fan system
operation
•Offers optional 408V power distribution with a choice of 208V
or 277V 3-phase power supplies
Reliable and Serviceable
•No single point of failures with fault tolerance capabilities
•Provides easy access to all critical components which are also
hot-swappable
•Offers redundant power, cooling and management subsystems
with failover capabilities
•Includes built-in multi-generation configuration software
management
Flexible and Scalable Configurations
•Based on the Intel® Xeon® Processor E5 family
•Offers 2 and 4 sockets CPUs or hybrid processing based platforms
•Offers a variety of interconnects: 3D Torus/Fat Tree, single/dual
rail, QDR/FDR, IB/GigE
•Supports a wide range of parallel Storage File Systems
Manageable
•Integrated HPC software stack
•Includes Advanced Cluster Engine™ (ACE) management software
•Offers remote system control manager with CLI and GUI
•Supports multi Linux OS environments
•Manages heterogeneous nodes with different OS stacks
•Offers applications middleware such as message passing libraries,
compilers, debuggers, and performance tools
•Includes network, server, cluster, and storage management
•Monitors fine-grain system power and temperature
•Provides version control and the ability to rollback changes
•Integrates job schedulers such as Grid Engine, SLURM, and PBS Pro
End-To-End Supercomputer Solutions
•Complete design, integration, testing and delivery
•On-site professional installation services and/or customized
services
Cray Xtreme-Cool Supercomputer Specifications
Architecture
Processor
Memory
Interconnect
External I/O
Interface
System
Administration
File System
Liquid-cooled cluster architecture offers 80 nodes per rack cabinet
Based on 8 cores 64-bit Intel® Xeon® processor E5 family
Optional hybrid processing configurations with Intel® Xeon Phi™
coprocessors or GPU computing accelerators
32, 64 or 128GB registered ECC DDR3 SDRAM per compute node
Up to 8GB 8 channels ECC GDDR5 memory per coprocessor node
- 1, 10 or 40 GE Gigabit Ethernet
- QDR or FDR InfiniBand with Connect X3 or
True Scale Host Channel Adapters
- Options for single or dual-rail Fat Tree or 3D Torus
- Advanced Cluster Engine (ACE): Complete lights-out remote
management capability
- Graphical and command line system administration
- System software version rollback capability
- Redundant management servers with automatic
load balancing and failover
- Automatic discovery and status reporting of interconnect,
server, and storage hardware
- Partition a cluster into multiple logical clusters each capable
of hosting a unique software stack
- Integrated job scheduling and management
- Enable the export, import of system configurations and images
- Ability to detect HW, fabric topology configuration errors
- Remote server control (Power On/Off, Cycle) and remote
server initialization (Reset, Reboot, Shut Down)
- Scalable fast diskless booting for large node systems and
root file systems for diskless nodes
- Multiple global storage configurations
NFS, Lustre® and Panasas® PanFS available as Global File System
Total Energy: 2.007MW
Savings: 407kW
+
Potential to Recover
up to 1520kW
Disk Storage Full line of FC-attached disk arrays with support for FC,
SATA disk drives and SSDs
Reliable, - Redundant power, cooling, and management servers with
Available, failover capabilities
Serviceable - Redundant management networks (InfiniBand, GbE & 10GbE)
(RAS) with failover
- Redundant root file system
- Built-In multi-generation configuration software management
- All critical components are easily accessible and hot-swappable
OS RedHat, Suse or Cent OS
Compilers, - Open MPI, MVAPICH2 or Intel® MPI Libraries
Libraries - PGI, PathScale, Intel® Cluster Toolkit compilers, CUDA,
and Tools CUDA C/C++/Fortran,
- OpenCL, DirectCompute Toolkits, GNU, TotalView,
OFED programming tools
Power - Up to 28KW per cabinet depending on configuration
- Optional 480V power distribution with a choice of 208V or
277V 3-phase power supplies
Cooling - Liquid-cooled up to 80% heat capture
Features - Offers low pressure secondary loop that is completely isolated from
the primary data center liquid loop
- Field serviceable cooling kits with integrated pressure and leak
detection remote monitoring
Cabinet 82.40”(2093mm) H x 23.62”(600mm) W x 59.06”(1500mm) D
Dimensions Standard 42U/19” rack cabinet
Cabinet Weight 295 lbs. rack + 100 lbs. CDU
Support & Up-front consulting, project management, system integration,
Services system installation, operations training
Dry Cooler 36kW
IT Power
1.9MW
Recovery
Potential
1.52MW
CRAC 19kW
Chiller Plant
~137 Tons
52kW
50˚F (10˚C)
59˚F (15˚C)
104˚F (40˚C)
138˚F (58.9˚C)
Cray Inc. • 901 Fifth Avenue, Suite 1000 • Seattle, WA 98164 • Tel: 206.701.2000 • Fax: 206.701.2500 • www.cray.com
©2012 Cray Inc. All rights reserved. Specifications subject to change without notice. Cray is a registered trademark, and the Cray logo, Cray Xtreme-Cool, Cray Xtreme-X, Advanced Cluster Engine are trademarks of Cray Inc.
Intel, the Intel logo, Xeon, and Xeon Inside are trademarks or registered trademarks of Intel Corporation in the U.S. and other countries. All other trademarks mentioned herein are the properties of their respective owners. 20121205JRC