User:JSharp/Verax

From RCS Wiki
Jump to navigation Jump to search

T2/x86 FSC Heterogeneous Cluster (verax@kraftwerk)

verax cluster
Verax Central Processor Complex

Includes Primary User Interface/System Element (in the IBM sense) components. A work in progress, incomplete - in general, ask on the discussion page.

Primary Purpose

  • Heterogeneous Distributed OS Development
  • K8s development
  • Real-time distributed raytracing, nonvolatile ram enabled persistent world simulation
  • Data Processing/Numerical Analysis Workloads for Academic Users
  • Firmware Integration and Development
  • Software Build Services for OpenPOWER Compatibility/Performance Related Open Source Projects
  • Full-control/debugging environment for low-level systems testing and verification including hardware/bus interposers (digital/analogue capture) for qualified local/remote users
  • Sub-ambient Cooling and Overclocking/Power Delivery Testing

Components

Incomplete list, acquisition status to be added (though the majority of components have been acquired).

  • Infrastructure
    • 1x CyberPower CR24U11001 24U Rack Enclosure
    • 3x TCG TGC-H4-650 Chassis 4U 650mm Depth w/o PSU
    • 1x Thermaltake Core W200 Chassis (Dual System, 1x RCS T2/1x x86)
  • Power Routing and Management
    • 2x APC SMX3000HV Smart-UPS 3000VA (4U)
    • 2x CyberPower PDU20MHVIEC10AT Metered Auto-transfer Switch (1U)
  • Power Conversion
    • 5x Corsair AX1500i Power Supply
  • Data Routing/Transfer
    • 1x Arista DCS-7050QX-32 (QSFP+) (1U)
    • 1x Cisco SG-500X-24-K9 (1U)
    • 10x Mellanox MCX314A-BCBT ConnectX-3 Ethernet Adapters (Dual Port)
  • Data Storage Elements
    • 30x Seagate BarraCuda ST8000DM004 8TB SATA HDDs (coldline)
    • 10x Samsung EVO PRO 860 1TB SATA SSDs (warmline)
    • 5x LSI 9300-8i SATA/SAS HBA
    • 8x Intel 900P (4x PCI Express/4x OCuLink) NVMe 480GB (hotline)
    • 32x Samsung M393A2G40EB2 16GB DDR4-2667 ECC DRAM (online - POWER) (64x once tested)
    • 8x Micron MT36KSF2G72PZ-1G6E1FE 16GB DDR3-1600 ECC DRAM (online - x86) (16x once memory controller patched, known coreboot issues)
  • Data Processing Elements/Backplane
    • 4x Raptor Computing Systems T2P9D01 Talos II Mainboard
    • 1x ASUS KGPE-D16 Mainboard
    • 2x AMD Opteron 6378 CPU
    • 8x IBM POWER9 CPU (18-Core)
    • 6x AMD Radeon Pro WX 7100