Supermicro is one of the top vendors of high-performing computing hardware products. This includes its CPU, GPU, and FPGA products. As a result, Supermicro delivered substantially superior performance than its competition with its HPC cluster solutions for deep learning, neural network training, and image processing applications.
Today’s application demands are driving greater compute capability, met by supermicro’s HPC clusters designed to help solve challenges in Internet security and Molecular Biology research.
Supermicro’s HPC cluster solutions are the best and most affordable way to achieve superior results for your next project; we will review the Supermicro HPC infrastructure, obtain our data in the cloud, and analyze it.
We will be working with Google Cloud Platform (GCP) and using their Cloud Machine Learning Engine (Cloud ML Engine), where we can utilize their machine learning capabilities and get hands-on experience in this domain. In this project, we will use the Titanic dataset, which can be found here.
In this project, we will use two Supermicro systems, the SYS-5038A-I and SYS-5039MC-HTRF. The network interface controller (NIC) used in both systems is the Intel X540 Dual Port 10GBASE-T Ethernet Adapter. Both methods have a 2x Intel Xeon E5-2637 v3 processor with two cores per CPU.
Both Supermicro systems have high-performance storage devices, including NVMe SSDs and hard drives for each design. Configuration of these systems varies according to the customer’s requirement but always meets the most demanding expectations in both performance and capacity.
Windows Server 2008 R2, Windows Server 2012, and Windows Server 2016 are all supported by the hyper-threading technology; however, the Hyper-V hyper-threaded environment does not work.
Hyper-V support is available for the following major operating systems:
Hardware support for virtualization is also included in both systems. The System Chassis has two Intel Xeon E5-2637 v3 processors installed with eight physical cores, all active for virtualization use, and two Intel PCIe slots.
Each server has a maximum 192TB raw storage capacity in Ceph object storage. The OS-level virtualization of Supermicro has been designed to be highly compatible with VMware and Hyper-V, which we will look into later in this article.
The Supermicro SYS-5038A-I supports up to 4x Ethernet ports with the 10GbE Intel X540 Dual Port 10GBASE-T Ethernet Adapter, while the SYS-5039MC-HTRF supports up to 5x Ethernet ports with the same adapter. In addition, both systems have NIC failover capability. All peripherals connected to Supermicro products are also supported in live migration, which is essential for high availability.
The network configuration on these HPC cluster systems is an 8-port 10GbE Intel X540 chipset with a total bandwidth of 80Gbps. The 10GbE can be bonded to increase bandwidth, which allows for convenient interconnection of different clusters in different sites for more flexibility. The two systems also support many VLAN and bonding options for high availability.
In addition to the 10GbE network, the systems can also be used in a virtualized environment via 10GBase-T. For high availability purposes, both servers support NFS and CIFS protocols for easy file sharing. They also have an IPMI interface for management use and serial consoles for better efficiency.
In addition to the above hardware configuration, the two systems are also equipped with dual hot-swap PSUs. They are high-performance, reliable, and can be quickly swapped in case of a power failure.
The power supply is built using thermally efficient components while providing sufficient output capacity. The system casing is constructed using high-grade metal materials to ensure optimal heat dissipation, reliability, and stability.