Recommended Hardware Configuration for SQream DB

Chassis

CPU

RAM

GPU

Chassis capable of housing 2-4 NVIDIA® Tesla GPUs Dual IBM POWER9 or Dual Intel Xeon Gold processor 128 GB RAM per logical GPU NVIDIA Tesla GPU
At least 2 NVIDIA® Tesla series GPUs
Examples: 
IBM AC922 POWER9, Dell R740, Dell R940xa, NVIDIA® DGX-1, Cisco UCS C240 M4, HP DL380g10, Lenovo System x3650 M5, Supermicro SYS-2028U / 6028U
Recommended:
IBM Power9 with (16C) or Intel Xeon Gold 6150 (18C)
Recommended: 
512GB - 1024GB
Recommended: 
2, 4 or 6 NVIDIA® Tesla V100

SQream – NVIDIA® Tesla V100 PCIe
  • Server capable of housing 2-4 NVIDIA® Tesla GPUs
  • PCIe 32GB/s bandwidth
  • 1 Tesla V100 GPU per 2 users
  • SQream recommends about 256GB of RAM per physical GPU

Examples: IBM AC922 POWER9, Dell R740, Dell R940xa, NVIDIA®  DGX-1




SQream – NVIDIA® Tesla V100 NVLink
  • Server capable of housing 2-4 NVIDIA® Tesla GPUs
  • NVIDIA® NVLink 300GB/s bandwidth
  • 1 Tesla V100 GPU per 2 users
  • SQream recommends about 256GB of RAM per physical GPU

Examples: IBM AC922 POWER9, Dell EMC PowerEdge C4130


Suitable for - 50 Users, Up to 3PB in just 10U (Networking Included)



Independent Scaling of Storage and Compute
SQream DB can scale to unlimited data sizes, many data consumers, or both. Because compute is decoupled from storage, scaling is available for just storage, just compute, or both.


SQream DB has a shared-data architecture that enables independent scaling of storage and compute.