Home CPU CLusters1

Saturday25 May 2019

C10K - CPU based Clusters

NOVATTE C10K CPU Cluster Appliance is APAC best preconfigured HPC system which brings plug-and-play simplicity for small to large deployments. Using unified NOVATTE architecture, Admin-friendly CPU Cluster can be easily scaled from small research-lab cluster to large-scale several hundreds TFlops Supercomputer without any degrade in performance.
novatte rack

Intel Cluster Ready
All NOVATTE CPU Cluster Appliances are officially certified by Intel for Intel Cluster Ready program. Be confident your cluster will work as it should, right away, without worrying if components will work together or not.

 

High Performance and Scalability
Being agile and first to market, NOVATTE uses the latest possible technologies to accelerate its HPC Clusters and bring performance to the maximum. Directly connecting computing and management nodes with storage through non-blocking 56Gbps FDR InfiniBand network, NOVATTE unifies system architecture to decrease protocol bottlenecks and allow NOVATTE HPC Cluster almost linear performance scalability of up to several PFLOPS. Unification of NOVATTE architecture makes it easy to convert default NFS storage file system to Lustre Parallel File System for maximum performance, scale and failover.

 

Great Value, Low TCO
Due to the shorter “Technology Vendor – NOVATTE – End User” value chain and thus lower overheads and costs, NOVATTE CPU Cluster Appliance outcompetes its closest competitors in price/performance to a great extent. Excellent investment protection, increased CPU utilization rates, high density and simplified cluster management delivers an attractive total cost of ownership while fully addressing the big data requirements of even the most compute-intensive technical applications.


c10kperf2
 

Exclusive benefits:

    - Turnkey pre-installed complete solution - switch it on and start working

    - High scalability and flexibility through unified system architecture

    - Intel Cluster Ready certified

    - Easy to use Admin-friendly intuitive cluster management system

    - Most competitive Price per GFlop in APAC

    - Highly competent instant support from the most experienced HPC engineers and practitioners in Asia

 

Sample configurations

   Base Quarter  Half Full Extension
   CPU Cluster base CPU Cluster quarter
CPU Cluster half
CPU Cluster full
CPU Cluster extention
Nodes 8 16 32 64 80
GFlop per system
(Rpeak)
2,995.2 5,657.6 10,982.4 21,632 26,956.8
kW per system 3.4 6.1 11.5 22.4 27.8
GFlops per kW 880.94 927.48 954.99 965.71 969.67
GB per Node 32 32 32 32 32
Storage NFS, optional: NLA - NOVATTE Lustre Appliance


All NOVATTE CPU Clusters in-a-box show Linpack efficiency of up to 88%.

HPC SW Stack

 

Specifications

Architecture
  • Separate data, monitoring and management networks
  • Redundant head nodes (HA) and provisioning nodes with both automatic and manual failover configurations
  • All critical components are easily accessible and hot-swappable
  • Diskless nodes (Linux HPC clusters)
  • Lustre data tiring (SSD NLA blocks for High IOPS demanding tasks and SAS NLA blocks for capacity demanding tasks)
Computational field
  • 8 cores 64-bit Intel® Xeon® processor E5 family
  • NVIDIA TESLA GPU accelerators
  • Intel® Xeon PhiTM coprocessors 
Memory
  • 32GB, 64GB and 128GB per compute node
  • DDR3-1600 1.35V low-voltage memory
  • DDR3-1866 1.5V memory
Networking/
Interconnect
  • 1GbE management network
  • 10 or 40GbE Gigabit Ethernet interconnect network
  • 40Gb/s FDR or 56Gb/s FDR InfiniBand interconnect network with ConnectX-3 or Intel® TrueScale single/dual port adapters
File System
  • Lustre
  • PanFS
  • NFS
Storage Systems
Power 
  • Auto 400V, 47-63 Hz 3-phase input power
  • Power consumption: 27.8 kW max
System Administration
  • Graphical or shell system administration and provisioning for both HPC Cluster and Lustre storage
  • Automatic interconnect, nodes and storage detection and monitoring
  • Cluster partitioning to smaller logical clusters running own software stack and own jobs
  • Automatic system software updates; rollback capabilities
  • Visual tools for comprehensive set of software and hardware metrics monitoring
  • Redundant head nodes with automated failover and load balancing
  • Integrated job schedulers
  • GPU and MIC management
  • ScaleMP management
  • Graphical monitoring and provisioning of Parallel File System, storage



JA Minisite

2857.orig.q75.o0 - Copy IntelTPP amd    qctlogo e   Mellanox APAC Partner   1