HDPOD System 300 Solution Design Architecture Overview

IRON HDPOD is designed to be a Hortonworks Hardoop HDFS powered Big Data storage platform for running enterprise data analytics environments, the design offers:

  • Rack-Level Resiliency:Resiliency is built into the design at multiple levels to deal with hardware and software failures; all components have N+1. The degree of resiliency is adaptable for mitigating power, network, storage and compute outages.
  • Modular, Scalable Storage Architecture:Balanced compute CPU, memory, network throughput and storage IOPS and capacity per rack. Multiple racks can be networked together to build scale-out modular storage infrastructure
  • Flexible Data Node Configurations:For Hadoop workloads, the storage to compute ratios vary from application to application; HDPOD offers multiple building block choices in 2U, 1U and half-U form factors, as well as Low Power consumption CPU options.
  • High Return-on-Investment (ROI):Best-in-class standardized components to provide maximum performance leveraging synergies between components that are based on open standards, reliable and scalable, allowing organizations to save on cost and help grow with business needs.

IRON HDPOD Commodity Systems: Hardware Configurations Options

Hardware Components & Specifications H220SH H320SF H320S-P2 H320-P4
Network Fabric
• Aggregate/ Data Storage Switch: 1U, 36x 40/56GbE QSFP Ports Ethernet Switch 2 2 4 8
• Management Switch: 1U, 44x 1GbE RJ45, 4x SFP Ports Ethernet Switch 1 1 2 4
 
Cluster Management Server Appliance
• Management  Node Server: 1U, IRON iServer Single Node, 2x 10GbE & 1-IPMI Ports 1 1 1 1
           
Name Nodes for Hadoop
• Nodes (Master Server) 2U, IRON iServer Single Node, 2x 10GbE & 1-IPMI Ports 3 3 3 3
• CPU: Intel Dual Socket, 12 Core, 2.6Ghz CPU (Total 24 Cores/Node)        
• Memory (GB): 96, 128, 192, 256, 384 & 512 Options (256GB default)        
• Disk Drives Number of Disks – 10x; 900GB 10K RPM  (per Node)        
     Three nodes: 1x Primary named node, 1x Secondary named nodes (+HBase Master, Hive Server), 1x Job Tracker node
Data Storage Nodes for Hadoop HDFS        
• Blade Enclosures: 4U, IRON iServer, Quad Node Blade Chassis (4 blades) 4 8 17 24
• Nodes ( Server Blades): 4 Server Nodes/Chassis, 2x 40Gbe LAN & 1-IPMI Ports 16 32 68 140
• Processors (CPU/Cores): Dual Socket, 6 Core, 2.0Ghz CPU (Total 12 Cores/Blade) 32/192 64/384 136/816 280/1,680
• Memory (GB): 96, 128, 192, 256, 384 & 512 Options (128GB default) 2,048 4,096 8,704 17,920
• Disk Drives Number of Disks – 6,8 or 12 per Node (6 default) 96 192 408 840
    Total Disk  Capacity (Raw*) 1TB, 2TB, 3TB & 6TB SAS  (4TB default) 384 768 1,632 3,360
*Capacities listed uncompressed (3x compression assumed typical)  without Compression
    Total Disk  Capacity (Usable)    Usable mirrored capacity (appr.) 192 384 816 1,680
    Total Disk  Capacity (Usable)    Usable Triple mirrored capacity (appr.) 128 256 644 1,120
 
Additional Hardware and Software
• Rack IRON 42U Rack, Redundant PDUs 1 1 2 4
• KVM IRON Keyboard/Mouse/Video Display Console, Optional        
• Software (System) Linux Operating System, 64bit        
  Microsoft Windows Server 2012 R2 Optional        
• Software (Add-On), Optional Data Protection for Volume & MapReduce layers        
  Dynamic Data Set management, metadata, and data lineage        
 
Untitled Document