Rate This Document
Findability
Accuracy
Completeness
Readability

Typical Configuration

Table 1 describes the configuration of each component in offline analysis scenarios.

Table 1 Typical configuration for offline big data analysis

Node Type

Typical Configuration

Node Quantity

Quantity Calculation

Management node

Two-socket rack server, two Huawei Kunpeng 916 or Kunpeng 920 processors, 128 GB or larger memory, six 600 GB or larger 2.5-inch SAS drives, 1 GB LSI RAID 0/1 card (more than three RAID 1 arrays are supported), two 10GE network ports, two GE network ports (the two network ports are bonded and connected to two access switches), and independent power supply

2

Active/standby architecture: one active node + one standby node

Controller node

Two-socket rack server, two Huawei Kunpeng 916 or Kunpeng 920 processors, 256 GB or larger memory, ten 600 GB or larger 2.5-inch SAS drives, 1 GB LSI RAID 0/1 card (more than five RAID 1 arrays are supported), two 10GE network ports, two GE network ports (the two network ports are bonded and connected to two access switches), and independent power supply

3/5/9/11

  • Cluster scale 30–100: three sets
  • Cluster scale 100–500: five sets
  • Cluster scale 500–2000: nine sets
  • Cluster scale 2000–5000: 11 sets

Management and controller node (hybrid deployment)

Two-socket rack server, two Huawei Kunpeng 916 or Kunpeng 920 processors, 256 GB or larger memory, twelve 600 GB or larger 2.5-inch SAS drives, 1 GB LSI RAID 0/1 card (more than six RAID 1 arrays are supported), two 10GE network ports, two GE network ports (the two network ports are bonded and connected to two access switches), and independent power supply

3

Cluster scale 3–30

Data node

Two-socket server, two Huawei Kunpeng 916 or Kunpeng 920 processors, 256 GB or larger memory, two 600 GB 2.5-inch SAS drives, twelve 4 TB or larger 3.5-inch SATA drives, 1 GB LSI RAID 0/1 card (more than one RAID 1 array is supported), two 10GE network ports, and two GE network ports (the two network ports are bonded and connected to two access switches), and independent power supply

Calculated based on the data volume

The calculation formula is as follows: Number of nodes = Planned data volume x 1.5 (data expansion rate) x 1 (data compression rate) x 3 (three copies)/0.8 (drive utilization)/0.9 (drive numeral system conversion)/(12 (number of drives) x 4 TB (drive capacity))