Updated on 2024-04-15 GMT+08:00

High-Performance Computing ECSs

Overview

  • The processor and memory ratio of an Hc2 (high-performance computing second-generation) ECS is 1:2 or 1:4. Each vCPU corresponds to the hyperthreading of an Intel® Xeon® Scalable processor core. Hc2 ECSs can be used for high-performance computing services. They provide a large number of parallel computing resources and high-performance infrastructure services to meet the requirements of high-performance computing and massive storage and ensure rendering efficiency. HC2 ECSs are frequently used in the following scenarios:
    • Computing and storage systems for genetic engineering, games, animations, biopharmaceuticals, and scientific computing
    • Public rendering platforms for renderfarms and animation and film bases; other rendering platforms for movies and videos
    • High-performance frontend clusters, web servers, high-performance science and engineering applications, advertisements, video coding, and distributed analysis
  • H2 ECSs are designed to meet high-end computational needs, such as molecular modeling and computational fluid dynamics. In addition to the substantial CPU power, the H2 ECSs offer diverse options for low-latency RDMA networking using EDR InfiniBand NICs to support memory-intensive computational requirements.
  • Hl1 ECSs feature large memory capacity. They allow ECSs to interconnect with each other using 100 Gbit/s RDMA InfiniBand NICs and support 56 Gbit shared high I/O storage.
  • H3 ECSs use high-performance Intel® Xeon® Scalable processors. Each vCPU corresponds to the hyper thread of an Intel® Xeon® Scalable processor core, providing stable computing capabilities. H3 ECSs are suitable for high-performance computing services. In addition, such ECSs use latest-generation network acceleration engines and DPDK rapid packet processing mechanism to provide high network performance.

Specifications

Table 1 High-performance computing ECS specifications

Type

vCPUs

Memory (GiB)

Flavor

Virtualization

Hc2

2

4

hc2.large.2

KVM

4

8

hc2.xlarge.2

KVM

8

16

hc2.2xlarge.2

KVM

16

32

hc2.4xlarge.2

KVM

32

64

hc2.8xlarge.2

KVM

2

8

hc2.large.4

KVM

4

16

hc2.xlarge.4

KVM

8

32

hc2.2xlarge.4

KVM

16

64

hc2.4xlarge.4

KVM

32

128

hc2.8xlarge.4

KVM

Hl1

32

256

hl1.8xlarge.8

KVM

H2

16

128

h2.4xlarge.8

KVM

16

256

h2.4xlarge.16

KVM

H3

2

4

h3.large.2

KVM

4

8

h3.xlarge.2

KVM

8

16

h3.2large.2

KVM

12

24

h3.3xlarge.2

KVM

16

32

h3.4large.2

KVM

24

48

h3.6xlarge.2

KVM

32

64

h3.8large.2

KVM

2

8

h3.large.4

KVM

4

16

h3.xlarge.4

KVM

8

32

h3.2large.4

KVM

12

48

h3.3xlarge.4

KVM

16

64

h3.4large.4

KVM

24

96

h3.6xlarge.4

KVM

32

128

h3.8large.4

KVM

Scenarios

  • Applications

    H2 and Hl1: High-performance computing (HPC), big data, and Artificial Intelligence (AI)

  • Application scenarios

    H2 and Hl1

    • High-performance hardware: The ratio of memory to vCPU is 8:1, and a large number of multi-thread physical CPUs are available to provide high-performance storage I/O and high-throughput network connections.
    • Designed for HPC clusters: Multiple Hl1 ECSs can be clustered to install scalable, clustered file system, such as Lustre. HPC applications running on H2 ECSs can read and modify the data stored in the ECSs.
    • RDMA network connection: Same as H2 ECSs, Hl1 ECSs also offer RDMA network using EDR 100 Gbit/s InfiniBand NICs. Hl1 ECSs can communicate with H2 ECSs with RDMA protocol. In addition, Hl1 ECSs can access EVS disks over the RDMA protocol, which allows up to 56 Gbit/s storage bandwidth.
  • Application scenarios

    H2 and Hl1 ECSs provide computing capabilities for clusters with a large memory, good connectivity between nodes, and high storage I/O. The typical application scenarios include HPC, big data, and AI. In HPC solution, Hl1 ECSs are perfectly suited for the Lustre parallel distributed file system, generally used for large-scale cluster computing.

    For example, in HPC scenario, H2 ECSs can be used as compute nodes, and Hl1 ECSs can be used as storage nodes.

Features

High-performance computing ECSs have the following features:

  • They provide large memory capacity and more processor cores than other types of ECSs.
  • They offer up to 32 vCPUs.
  • H2 and Hl1 ECSs use InfiniBand NICs that provide a bandwidth of 100 Gbit/s.
  • Hl1 ECSs can use the following types of EVS disks as system disk and data disk:
    • High I/O (performance-optimized I)
    • Ultra-high I/O (latency-optimized)
  • Hl1 ECSs support 56 Gbit shared high I/O storage.

    To support 56 Gbit shared high I/O storage, you only need to attach high I/O (performance-optimized I) or ultra-high I/O (latency-optimized) EVS disks to target Hl1 ECSs.

Notes on Using H2 ECSs

  • H2 ECSs do not support OS reinstallation or change.
  • H2 ECSs do not support specifications modification.
  • H2 ECSs do not support cold migration, live migration, or HA.
  • H2 ECSs support the following OSs:
    • For public images:
      • CentOS 7.3 64bit
      • SUSE Linux Enterprise Server 11 SP4 64bit
      • SUSE Linux Enterprise Server 12 SP2 64bit
    • For private images:
      • CentOS 6.5 64bit
      • CentOS 7.2 64bit
      • CentOS 7.3 64bit
      • SUSE Linux Enterprise Server 11 SP4 64bit
      • SUSE Linux Enterprise Server 12 SP2 64bit
      • Red Hat Enterprise Linux 7.2 64bit
      • Red Hat Enterprise Linux 7.3 64bit
  • H2 ECSs use InfiniBand NICs that provide a bandwidth of 100 Gbit/s.
  • Each H2 ECS uses one PCIe 3.2 TB SSD card for temporary local storage.
  • If an H2 ECS is created using a private image, install an InfiniBand NIC driver on the ECS after the ECS creation following the instructions provided by Mellanox. Download the required version (4.2-1.0.0.0) of InfiniBand NIC driver from the official Mellanox website and install the driver by following the instructions provided by Mellanox.
  • For SUSE H2 ECSs, if IP over InfiniBand (IPoIB) is required, you must manually configure an IP address for the InfiniBand NIC after installing the InfiniBand driver. For details, see "How Can I Manually Configure an IP Address for an InfiniBand NIC?" in the Elastic Cloud Server User Guide.
  • After you delete an H2 ECS, the data stored in SSDs is automatically cleared. Therefore, do not store persistence data into SSDs during ECS running.

Notes on Using Hl1 ECSs

  • Hl1 ECSs only support the attachment of high I/O (performance-optimized I) and ultra-high I/O (latency-optimized) EVS disks.

    To support 56 Gbit shared high I/O storage, you only need to attach high I/O (performance-optimized I) or ultra-high I/O (latency-optimized) EVS disks to target Hl1 ECSs.

  • Hl1 ECSs do not support specifications modification.
  • Hl1 ECSs use InfiniBand NICs that provide a bandwidth of 100 Gbit/s.
  • Hl1 ECSs created using a private image must have the InfiniBand NIC driver installed. Download the required version (4.2-1.0.0.0) of InfiniBand NIC driver from the official Mellanox website and install the driver by following the instructions provided by Mellanox.
    • InfiniBand NIC type: Mellanox Technologies ConnectX-4 Infiniband HBA (MCX455A-ECAT)
    • Mellanox official website: http://www.mellanox.com/
  • For SUSE Hl1 ECSs, if IP over InfiniBand (IPoIB) is required, you must manually configure an IP address for the InfiniBand NIC after installing the InfiniBand driver. For details, see "How Can I Manually Configure an IP Address for an InfiniBand NIC?" in the Elastic Cloud Server User Guide.
  • Hl1 ECSs support the following OSs:
    • For public images:
      • CentOS 7.3 64bit
      • SUSE Linux Enterprise Server 11 SP4 64bit
      • SUSE Linux Enterprise Server 12 SP2 64bit
    • For private images:
      • CentOS 6.5 64bit
      • CentOS 7.2 64bit
      • CentOS 7.3 64bit
      • SUSE Linux Enterprise Server 11 SP4 64bit
      • SUSE Linux Enterprise Server 12 SP2 64bit
      • Red Hat Enterprise Linux 7.2 64bit
      • Red Hat Enterprise Linux 7.3 64bit