|
Two units of the shared-memory system, HPE Superdome Flex (Intel Xeon 3.1GHz/32 sockets (576 cores)/24TB memory), have been installed.
Hardware
The HPE Superdome Flex (hereafter referred to as SDF) is a system equipped with 18-core Intel 64-bit Xeon processor G6254 3.1GHz on 32 sockets (576 cores), and two units have been installed.
It also includes 24TB of main memory.
A major feature of SDF is that a single 24TB memory space can be accessed from one operating system.
Because it adopts a distributed shared memory architecture (ccNUMA), each processor has its own local memory (distributed memory), and all node memories are logically combined into shared memory. This allows all physical memory of the system to be referenced as a single memory space. As a result, users can develop programs without needing to consider the size of the memory they use.
In SDF, two systems of identical configuration have been installed. Therefore, even if one experiences a failure or undergoes scheduled maintenance, the other system can continue to provide service, resulting in a configuration that enhances availability.
| Superdome Flex |
|---|
| CPU Model | Intel Xeon Gold 6254 (Cascade Lake) |
| Base Clock Frequency | 3.1GHz |
| L3 Cache | 24.8MB |
| Number of Cores per Node | 576 |
| Theoretical CPU Performance per Node | 57.1TFLOPS |
| Main Memory per Node (Transfer Rate) | DDR4-2933 24TB (2933MT/s) |
| Internal Disk per Node | 1.2TB SAS x2 (RAID1 configuration) |
| NVMe SSD Disk per Node | 51.2TB |
| GPU per Node | None |
| OS | Red Hat Enterprise Linux 8.7 |
| Interconnect | Between sockets:
Superdome Flex Grid (bidirectional 13.3GB/s) |
| Total Number of Nodes | 2 |
| Total Theoretical CPU Performance | 114.2TFLOPS |
| Batch Queue | SDF |
Software
The operating system running on SDF is Red Hat Enterprise Linux (RHEL) provided by Red Hat.
Security updates and technical support necessary for long-term stable operation of large-scale, mission-critical systems are provided.
SDF includes compilers, libraries, and toolsets that support automatic parallelization to easily harness the system's capabilities.
Tools for massive data analysis, such as genome assembly requiring large memory and metagenome analysis, are provided.
Additionally, various computational chemistry applications are available to perform quantum chemistry calculations and molecular dynamics simulations for large-scale systems that would otherwise lack sufficient computational resources on a cluster.
Related Links
HPE Superdome Flex
Red Hat Enterprise Linux
| 
