Cluster node specifications - Elastic High Performance Computing

This topic describes the recommended specifications for E-HPC and their common scenarios.

E-HPC supports all ECS instance types. This topic recommends several instance types that are well-suited for the High Performance Computing (HPC) features of E-HPC. Choose an instance type based on your scenario. For more information about ECS instance types, see Instance family and Best practices for instance type selection.

ECS Bare Metal Instances and Super Computing Cluster (SCC) Instances

The following instances have no virtualization overhead. They provide VPC network isolation and enable direct access to hardware resources. Super Computing Cluster (SCC) instances also provide low-latency Remote Direct Memory Access (RDMA) network interconnects to support the parallel computing requirements of E-HPC.

SCC instances

Instance Type	Instance type	Compute configuration	Network configuration	Storage configuration	Recommended scenarios
High clock speed + Low-latency network	ecs.scchfg6.20xlarge	Processor: 3.1 GHz Intel® Xeon® Platinum 8269 (Cascade Lake) vCPUs: 80 Memory: 384 GiB GPU: None	Ethernet: 30 Gbit/s RDMA: 50 Gbit/s	Ultra disks Enterprise SSDs (ESSDs) Standard SSDs	Ideal for services that require high clock speeds and parallel computing across multiple machines, such as simulation and manufacturing, life sciences, and weather forecasting.
	ecs.scch5s.16xlarge	Processor: 3.1 GHz Intel® Xeon® Platinum 8269 (Cascade Lake) vCPUs: 64 Memory: 192 GiB GPU: None	Ethernet: 30 Gbit/s RDMA: 50 Gbit/s	Ultra disks ESSDs Standard SSDs
	ecs.scch5.16xlarge	Processor: 3.1 GHz Intel® Xeon® Platinum 8269 (Cascade Lake) vCPUs: 64 Memory: 192 GiB GPU: None	Ethernet: 10 Gbit/s RDMA: 50 Gbit/s	Ultra disks Standard SSDs	Ideal for services that require high clock speeds and parallel computing across multiple machines, such as simulation and manufacturing, life sciences, and weather forecasting. ESSDs are not supported. If you need high-speed local disk storage, use scchfg6 or scch5s.
General-purpose + Low-latency network	ecs.sccg5.24xlarge	Processor: 2.5 GHz Intel® Xeon® Gold 6149 (Skylake) vCPUs: 96 Memory: 384 GiB GPU: None	Ethernet: 10 Gbit/s RDMA: 50 Gbit/s	Ultra disks Standard SSDs	For services that require parallel computing across multiple machines, such as simulation and manufacturing. ESSDs are not supported. If you need high-speed local disk storage, use scchfg6 or scch5s.
GPU-accelerated + Low-latency network	ecs.sccgn6.24xlarge	Processor: 2.5 GHz Intel® Xeon® Gold 6149 (Skylake) vCPUs: 96 Memory: 384 GiB GPU: 8 × NVIDIA Tesla V100	Ethernet: 30 Gbit/s RDMA: 50 Gbit/s	Ultra Disk ESSDs Standard SSDs	For single-machine or multi-machine parallel computing services that require GPUs, such as machine learning and large-scale molecular dynamics.

ECS Bare Metal Instances

Instance type	Instance type	Compute configuration	Network configuration	Storage configuration	Recommended scenarios
High clock speed	ecs.ebmhfc6.20xlarge	Processor: 3.1 GHz Intel® Xeon® Platinum 8269 (Cascade Lake) vCPUs: 80 Memory: 384 GiB GPU: None	Ethernet: 30 Gbit/s	Ultra disks Standard SSDs	For services that require high performance on a single machine.
High clock speed + Large memory	ecs.ebmhfr6.20xlarge	Processor: 3.1 GHz Intel® Xeon® Platinum 8269 (Cascade Lake) vCPUs: 80 Memory: 768 GiB GPU: None	Ethernet: 30 Gbit/s	Ultra disks Standard SSDs	For services that require high performance on a single machine.
General-purpose	ecs.ebmg6.26xlarge	Processor: 2.5 GHz Intel® Xeon® Gold 6149 (Skylake) vCPUs: 104 Memory: 384 GiB GPU: None	Ethernet: 30 Gbit/s	Ultra disks ESSDs Standard SSDs	Ideal for life sciences scenarios, such as bioinformatics, dynamics simulation, and new drug research and development.
General-purpose + Large memory	ecs.ebmr6.26xlarge	Processor: 2.5 GHz Intel® Xeon® Gold 6149 (Skylake) vCPUs: 104 Memory: 768 GiB GPU: None	Ethernet: 30 Gbit/s	Ultra disks ESSDs Standard SSDs	Ideal for scenarios such as electronic design automation (EDA) and life sciences.
GPU-accelerated	ecs.ebmgn6v.24xlarge	Processor: 2.5 GHz Intel® Xeon® Gold 6149 (Skylake) vCPUs: 96 Memory: 384 GiB GPU: 8 × NVIDIA Tesla V100	Ethernet: 30 Gbit/s RDMA: 50 Gbit/s	Ultra disks enterprise SSD Standard SSDs	For computing services that require GPUs, such as machine learning.

Instances from x86-based enterprise-level and heterogeneous computing families

The following instances have some virtualization overhead. You can choose an instance type that suits your scenario.

Instance type	Instance type	Compute configuration	Network configuration	Storage configuration	Recommended scenarios
General-purpose	ecs.g6.8xlarge	Processor: 2.5 GHz Intel® Xeon® Platinum 8269 (Cascade Lake) vCPUs: 32 Memory: 128 GiB GPU: None	Ethernet: 10 Gbit/s	ultra disk ESSDs Standard SSDs	For general, large-scale, concurrent services across multiple machines, such as animation rendering and life sciences.
High clock speed	ecs.hfg6.16xlarge	Processor: 3.1 GHz Intel® Xeon® Platinum 8269 (Cascade Lake) vCPUs: 64 Memory: 256 GiB GPU: None	Ethernet: 25 Gbit/s	Ultra disks enterprise SSD Standard SSDs	For computing services that require high clock speeds, such as life sciences applications.
Memory-optimized	ecs.re6.26xlarge	Processor: 2.5 GHz Intel® Xeon® Platinum 8269 (Cascade Lake) vCPUs: 104 Memory: 1536 GiB GPU: None	Ethernet: 16 Gbit/s	Ultra disks ESSDs Standard SSDs	For computing services that require large memory on a single machine, such as bioinformatics.
GPU-accelerated	ecs.gn6i-c24g1.6xlarge	Processor: 2.5 GHz Intel® Xeon® Gold 6149 (Skylake) vCPUs: 24 Memory: 93 GiB GPU: 1 × NVIDIA T4	Ethernet: 7.5 Gbit/s	Ultra Disk ESSDs Standard SSDs	For computing services that require GPUs.
GPU-accelerated	ecs.gn6i-c24g1.12xlarge	Processor: 2.5 GHz Intel® Xeon® Gold 6149 (Skylake) vCPUs: 48 Memory: 186 GiB GPU: 2 × NVIDIA T4	Ethernet: 15 Gbit/s	Ultra disks ESSDs Standard SSDs	For computing services that require GPUs.