Network Types Offered by Hyperstack
Explore Hyperstack's virtual machine networking solutions tailored to meet diverse workload demands and bandwidth requirements. These options include entry-level Ethernet, standard Ethernet, high-speed Ethernet with SR-IOV, and InfiniBand for Supercloud networks. This article also provides details on the network type associated with each of Hyperstack's GPU offerings.
In this article
- Entry-level Ethernet network
- Standard Ethernet network
- High-speed Ethernet networks with SR-IOV
- InfiniBand network for Supercloud
Entry-level Ethernet network
For our CPU-only and RTX GPU virtual machines, entry-level Ethernet networking provides a basic and efficient way to transfer data between connected devices. This networking standard is ideal for CPU-only and basic GPU-accelerated workloads, enabling efficient communication between virtual machines and external networks. With its reliability and affordability, entry-level Ethernet ensures virtual machines can efficiently exchange data, collaborate on tasks, and access remote resources, catering to various computational needs.
Use cases
- Virtual desktop infrastructure (VDI), 3D rendering, simulation, visualization, and more.
Supported GPUs
Standard Ethernet network
For workloads requiring faster data transfer speeds, Hyperstack offers the L40, A100, and H100 PCI-e GPU virtual machines. These machines are capable of achieving bandwidths ranging from 8 to 17 gigabits per second (Gbps), depending on thread count and MTU size.
Use cases
- LLM inference, LLM fine-tuning, image recognition, data analytics, and more.
Virtual Machine GPUs
PCI-e cards in the CANADA-1 region.
Supported GPUs
High-speed Ethernet network with SR-IOV
For workloads requiring higher bandwidth speeds, Hyperstack offers GPU virtual machines with high-speed Ethernet featuring SR-IOV (Single Root I/O Virtualization) technology. SR-IOV is a specification that allows a single physical network interface to be shared among multiple VMs, reducing latency and improving throughput. This capability is particularly useful in virtualized environments, enhancing the performance and efficiency of network traffic handling between virtual machines (VMs) and the physical network.
With SR-IOV technology, virtual machines in a network can achieve bandwidth speeds ranging from 37 to 350 gigabits per second (Gbps), depending on thread count and MTU size, with a peak bandwidth of 400 Gbps. Hyperstack GPUs compatible with high-speed networking include the A100 with NVLink, H100, and H100 with NVLink.
To learn more about high-speed networking, click here.
SR-IOV is available for contracted users upon request through our technical support team at [email protected].
Use cases
- LLM inference, LLM fine-tuning, data analytics, and more.
Supported GPUs
InfiniBand network for Supercloud
InfiniBand network allows clustering of multiple hosts with GPUDirect Remote Direct Memory Access (RDMA), enabling data transfer directly between their main memories while bypassing the processor, cache, and operating system. This results in extremely high throughput and very low latency, making it ideal for environments that require fast and efficient data transfer, such as high-performance computing (HPC) clusters and training large language models (LLMs).
A standout protocol within this framework is SHARP (Scalable Hierarchical Aggregation and Reduction Protocol), which optimizes collective communications, particularly in large parallel computing tasks, by offloading them to network hardware. By reducing data movement during these operations, SHARP significantly enhances efficiency. Additionally, InfiniBand includes Self-Healing Networks, which provide rapid recovery from network failures. This capability is crucial for maintaining the stability and reliability of large GPU clusters, ensuring minimal downtime and consistent performance even during network disruptions.
Use cases
- LLM inference, LLM fine-tuning, LLM foundational model training, Data Analytics, high-performance computing (HPC) for simulation, and scientific computing, and more.
Supercloud GPUs with InfiniBand
- NVIDIA H100 PCIe
- NVIDIA HGX B100 (coming soon)
- NVIDIA GB200 NVL72 (coming soon)
InfiniBand is exclusively available for Supercloud offerings.