NGX-CLUSTER

Based on your requirements, we will design and build a computer cluster with optimal performance, scalability, and reliability for scientific and technological calculations.

Configured for your computing needs

• We estimate the calculation accuracy, speed, and parallel processing required for the model to be calculated, and propose the optimal system configuration.
• We analyse the characteristics of your software engines, applications, and workloads to configure optimal computational resources.

Hardware Configuration

• CPU: We select the most optimized high-performance multi-core processors (such as Intel Xeon and AMD’s EPYC series).
• GPU: For artificial intelligence (machine learning, deep learning, etc.) applications, we recommend NVIDIA’s Tesla and Quadro series.
• Network: We identify bottlenecks in parallel calculations and streamline data communication between calculation nodes by using Ethernet bonding (redundancy and parallelization) and high-speed interconnects (InfiniBand).
• Computing nodes: We select machines from our NGX series and customize the required CPU and memory amount to maximize the computing power of the entire cluster.

Storage Design

• Taking into consideration data read/write speed and capacity, we build storage systems that combine high-speed SSDs and large-capacity HDDs.
• Implement proper RAID settings and backup strategies to ensure data backup and fault tolerance.

Redundancy and Scalability

• To increase system redundancy, we select redundant power supply units and network devices.
• We design your cluster to easily accommodate future expansions by integrating additional nodes or scaling out the cluster.

Software Compatibility, Job Management, Cluster Management

• We select and configure the appropriate OS (Linux, etc.) and drivers according to the scientific and technical calculation software and libraries to be used.
• A job scheduler (PBS, LFS, etc.) will be introduced to facilitate job management.
• In order to simplify management of the entire cluster, we will introduce integrated management tools, monitoring systems, virtualization systems, and hardware management interfaces, and build a managing system that allows the whole system to be managed centrally from the network.

Cloud Integration

• Cloud services can be used to flexibly expand computing resources as needed.
• Considering hybrid cloud architectures, we aim for seamless integration of cluster and cloud resources.