x86-64 instance profiles
When you provision IBM Cloud® Virtual Servers for Virtual Private Cloud, you can select from seven families of profiles: Balanced, Compute, Memory, Ultra High Memory, Very High Memory, Storage Optimized, and GPU.
A profile is a combination of instance attributes, such as the number of vCPUs, amount of RAM, network bandwidth, and default bandwidth allocation. The attributes define the size and capabilities of the virtual server instance that is provisioned. In the IBM Cloud console, you can select the most recently used profile or click View All Profiles to choose the profile that best fits your needs.
For more information about SAP profiles, see Intel Virtual Server certified profiles on VPC infrastructure for SAP HANA and Intel Virtual Server certified profiles on VPC infrastructure for SAP NetWeaver.
The following profile families are available when you provision a virtual server instance.
Family | Description |
---|---|
Balanced | Balanced profiles offer a core to RAM ratio that is best for midsize databases and common cloud applications with moderate traffic. |
Compute | Compute profiles offer a core to RAM ratio that is best for moderate to high web traffic workloads. Compute profiles are best for workloads with intensive CPU demands, such as high web traffic workloads, production batch processing, and front-end web servers. |
Memory | Memory profiles offer a core to RAM ratio that is best for memory caching and real-time analytics workloads. Memory profiles are best for memory intensive workloads, such as large caching workloads, intensive database applications, or in-memory analytics workloads. |
Very High Memory | Very High Memory profiles offer a core to RAM ratio of 1 vCPU to 14 GiB of RAM. This family is optimized for running small to medium in-memory databases and OLAP workloads, such as SAP BW/4 HANA. |
Ultra High Memory | Ultra High Memory profiles offer the most memory per core with 1 vCPU to 28 GiB of RAM. These profiles are optimized to run large in-memory databases and OLTP workloads, such as SAP S/4 HANA. |
GPU | GPU enabled profiles provide on-demand access to NVIDIA V100 and A100 GPUs to accelerate AI, high-performance computing, data science, and graphics workloads. |
Storage Optimized | Storage Optimized profiles offer temporary SSD instance storage disks at a ratio of 1 vCPU to 300 GB instance storage with a smaller price point per GB. These profiles are designed for storage-dense workloads and offer virtio interface type for attached disks. |
Confidential Compute | Confidential Compute-supported profiles use processor reserved memory called EPC (Enclave Page Cache) to encrypt application data. Processor reserved memory EPC maintains confidentiality and integrity. |
2nd generation profiles with instance storage and 2nd generation profiles with 64 or more vCPUs are deployed exclusively on the Intel®'s second-generation quad processor Xeon® Platinum 8260 Cascade Lake with 96 cores that are running at a base speed of 2.4 GHz and an all-core turbo frequency of 3.1 GHz or Intel®'s quad processor Xeon® Gold 6248 Cascade Lake with 80 cores that are running at a base speed of 2.5 GHz and an all-core turbo frequency of 3.1 GHz.
For more information about LinuxONE (s390x processor architecture) profiles, see s390x instance profiles.
Profiles with AMD-manufactured processors are available in the Toronto region.
Balanced
Balanced profiles provide a mix of performance and scalability for more common workloads. The Balanced profile family includes profiles with and without instance storage. The following table shows all Balanced profiles that are available for Intel® x86-64, and AMD x86-64 processors.
Balanced profiles with the bx2d prefix are available in the US South (Dallas), US East (Washington DC), Canada (Toronto), United Kingdom (London), Germany (Frankfurt), Spain (Madrid), Japan (Tokyo), Japan (Osaka), and Australia (Sydney) regions.
The following table shows all balance profiles that are available for x86-64.
Instance profile | vCPU / Cores | NUMA count | GiB RAM | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|---|
bx3d-2x10 | 2 / 1 | 1 | 10 | 4 | 1x65 |
bx3d-4x20 | 4 / 2 | 1 | 20 | 8 | 1x130 |
bx3d-8x40 | 8 / 4 | 1 | 40 | 16 | 1x260 |
bx3d-16x80 | 16 / 8 | 1 | 80 | 32 | 1x520 |
bx3d-24x120 | 24 / 12 | 1 | 120 | 48 | 1x780 |
bx3d-32x160 | 32 / 16 | 2 | 160 | 64 | 2x520 |
bx3d-48x240 | 48 / 24 | 2 | 240 | 96 | 2x780 |
bx3d-64x320 | 64 / 32 | 2 | 320 | 128 | 2x1024 |
bx3d-96x480 | 96 / 48 | 2 | 480 | 192 | 2x1560 |
bx3d-128x640 | 128 / 64 | 2 | 640 | 200 | 2x2080 |
bx3d-176x880 | 176 / 88 | 2 | 880 | 200 | 2x2860 |
Instance profile | vCPU / Cores | GiB RAM | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|
bx2-2x8 | 2 / 1 | 8 | 4 |
|
bx2d-2x8 | 2 / 1 | 8 | 4 | 1x75 |
bx2-4x16 | 4 / 2 | 16 | 8 |
|
bx2d-4x16 | 4 / 2 | 16 | 8 | 1x150 |
bx2-8x32 | 8 / 4 | 32 | 16 |
|
bx2d-8x32 | 8 / 4 | 32 | 16 | 1x300 |
bx2-16x64 | 16 / 8 | 64 | 32 |
|
bx2d-16x64 | 16 / 8 | 64 | 32 | 1x600 |
bx2-32x128 | 32 / 16 | 128 | 64 |
|
bx2d-32x128 | 32 / 16 | 128 | 64 | 2x600 |
bx2-48x192 | 48 / 24 | 192 | 80 |
|
bx2d-48x192 | 48 / 24 | 192 | 80 | 1x900 |
bx2-64x256 | 64 / 32 | 256 | 80 |
|
bx2d-64x256 | 64 / 32 | 256 | 80 | 2x1200 |
bx2-96x384 | 96 / 48 | 384 | 80 |
|
bx2d-96x384 | 96 / 48 | 384 | 80 | 2x1800 |
bx2-128x512 | 128 / 64 | 512 | 80 |
|
bx2d-128x512 | 128 / 64 | 512 | 80 | 2x2400 |
bx2a-2x8 | 2 / 1 | 8 | 2 |
|
bx2a-4x16 | 4 / 2 | 16 | 4 |
|
bx2a-8x32 | 8 / 4 | 32 | 8 |
|
bx2a-16x64 | 16 / 8 | 64 | 16 |
|
bx2a-32x128 | 32 / 16 | 128 | 32 |
|
bx2a-48x192 | 48 / 24 | 192 | 48 |
|
bx2a-64x256 | 64 / 32 | 256 | 64 |
|
bx2a-96x384 | 96 / 48 | 384 | 80 |
|
bx2a-128x512 | 128 / 64 | 512 | 80 |
|
bx2a-228x912 | 228 / 114 | 912 | 80 |
|
AMD-based virtual machines use AMD EPYC Milan processors. Compute capabilities are limited to AMD EPYC Rome capabilities.
Compute
Compute profiles are best for workloads with intensive CPU demands, such as high web traffic workloads, production batch processing, and front-end web servers. The Compute profile family includes profiles with and without instance storage. The following table shows all Compute profiles that are available for ® x86-64.
Compute profiles with the cx2d prefix are available in the US South (Dallas), US East (Washington DC), Canada (Toronto), United Kingdom (London), Germany (Frankfurt), Spain (Madrid), Japan (Tokyo), Japan (Osaka), and Australia (Sydney) regions.
The following table shows all compute profiles that are available for x86-64.
Instance profile | vCPU / Cores | NUMA count | GiB RAM | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|---|
cx3d-2x5 | 2 / 1 | 1 | 5 | 4 | 1x65 |
cx3d-4x10 | 4 / 2 | 1 | 10 | 8 | 1x130 |
cx3d-8x20 | 8 / 4 | 1 | 20 | 16 | 1x260 |
cx3d-16x40 | 16 / 8 | 1 | 40 | 32 | 1x520 |
cx3d-24x60 | 24 / 12 | 1 | 60 | 48 | 1x780 |
cx3d-32x80 | 32 / 16 | 2 | 80 | 64 | 2x520 |
cx3d-48x120 | 48 / 24 | 2 | 120 | 96 | 2x780 |
cx3d-64x160 | 64 / 32 | 2 | 160 | 128 | 2x1024 |
cx3d-96x240 | 96 / 48 | 2 | 240 | 192 | 2x1560 |
cx3d-128x320 | 128 / 64 | 2 | 320 | 200 | 2x2080 |
cx3d-176x440 | 176 / 88 | 2 | 440 | 200 | 2x2860 |
Instance profile | vCPU / Cores | GiB RAM | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|
cx2-2x4 | 2 / 1 | 4 | 4 |
|
cx2d-2x4 | 2 / 1 | 4 | 4 | 1x75 |
cx2-4x8 | 4 / 2 | 8 | 8 |
|
cx2d-4x8 | 4 / 2 | 8 | 8 | 1x150 |
cx2-8x16 | 8 / 4 | 16 | 16 |
|
cx2d-8x16 | 8 / 4 | 16 | 16 | 1x300 |
cx2-16x32 | 16 / 8 | 32 | 32 |
|
cx2d-16x32 | 16 / 8 | 32 | 32 | 1x600 |
cx2-32x64 | 32 / 16 | 64 | 64 |
|
cx2d-32x64 | 32 / 16 | 64 | 64 | 2x600 |
cx2-48x96 | 48 / 24 | 96 | 80 |
|
cx2d-48x96 | 48 / 24 | 96 | 80 | 1x900 |
cx2-64x128 | 64 / 32 | 128 | 80 |
|
cx2d-64x128 | 64 / 32 | 128 | 80 | 2x1200 |
cx2-96x192 | 96 / 48 | 192 | 80 |
|
cx2d-96x192 | 96 / 48 | 192 | 80 | 2x1800 |
cx2-128x256 | 128 / 64 | 256 | 80 |
|
cx2d-128x256 | 128 / 64 | 256 | 80 | 2x2400 |
Memory
Memory profiles are best for memory intensive workloads, such as large caching workloads, intensive database applications, or in-memory analytics workloads. The Memory profile family includes profiles with and without instance storage. The following table shows all Memory profiles that are available for Intel® x86-64.
Memory profiles with the mx2d prefix are available in the US South (Dallas), US East (Washington DC), Canada (Toronto), United Kingdom (London), Germany (Frankfurt), Spain (Madrid), Japan (Tokyo), Japan (Osaka), and Australia (Sydney) regions.
Instance profile | vCPU / Cores | NUMA count | GiB RAM | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|---|
mx3d-2x20 | 2 / 1 | 1 | 20 | 4 | 1x65 |
mx3d-4x40 | 4 / 2 | 1 | 40 | 8 | 1x130 |
mx3d-8x80 | 8 / 4 | 1 | 80 | 16 | 1x260 |
mx3d-16x160 | 16 / 8 | 1 | 160 | 32 | 1x520 |
mx3d-24x240 | 24 / 12 | 1 | 240 | 48 | 1x780 |
mx3d-32x320 | 32 / 16 | 2 | 320 | 64 | 2x520 |
mx3d-48x480 | 48 / 24 | 2 | 480 | 96 | 2x780 |
mx3d-64x640 | 64 / 32 | 2 | 640 | 128 | 2x1024 |
mx3d-96x960 | 96 / 48 | 2 | 960 | 192 | 2x1560 |
mx3d-128x1280 | 128 / 64 | 2 | 1280 | 200 | 2x2080 |
mx3d-176x1760 | 176 / 88 | 2 | 1760 | 200 | 2x2860 |
Instance profile | vCPU / Cores | GiB RAM | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|
mx2-2x16 | 2 / 1 | 16 | 4 |
|
mx2d-2x16 | 2 / 1 | 16 | 4 | 1x75 |
mx2-4x32 | 4 / 2 | 32 | 8 |
|
mx2d-4x32 | 4 / 2 | 32 | 8 | 1x150 |
mx2-8x64 | 8 / 4 | 64 | 16 |
|
mx2d-8x64 | 8 / 4 | 64 | 16 | 1x300 |
mx2-16x128 | 16 / 8 | 128 | 32 |
|
mx2d-16x128 | 16 / 8 | 128 | 32 | 1x600 |
mx2-32x256 | 32 / 16 | 256 | 64 |
|
mx2d-32x256 | 32 / 16 | 256 | 64 | 2x600 |
mx2-48x384 | 48 / 24 | 384 | 80 |
|
mx2d-48x384 | 48 / 24 | 384 | 80 | 2x900 |
mx2-64x512 | 64 / 32 | 512 | 80 |
|
mx2d-64x512 | 64 / 32 | 512 | 80 | 2x1200 |
mx2-96x768 | 96 / 48 | 768 | 80 |
|
mx2d-96x768 | 96 / 48 | 768 | 80 | 2x1800 |
mx2-128x1024 | 128 / 64 | 1024 | 80 |
|
mx2d-128x1024 | 128 / 64 | 1024 | 80 | 2x2400 |
Very High Memory
Very High Memory profiles offer 1 vCPU to 14 GiB of RAM to host small to medium-sized in-memory databases, OLAP services such as SAP NetWeaver, and other memory intensive applications. All Very High Memory profiles are provisioned with temporary SSD-backed instance storage at no additional charge. The following Very High Memory profiles are available on Intel® x86 processors.
The 3rd generation profile with the vx3d prefix is a beta feature that is available for evaluation and testing purposes. This profile is available only in the Toronto (ca-tor
) region to provision virtual server instances on 4th
Generation Intel® Xeon® Scalable processors, the Intel 8474C processor (previously code named Sapphire Rapids). For more information about the capabilities of the new profiles, see Next generation instance profiles.
- The vx2d profiles are on the Cascade Lake processors.
- The vx3d profiles are on the Sapphire Rapids processors.
Instance profile | vCPU | Cores | GiB RAM | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|---|
vx3d-2x32 | 2 | 1 | 32 | 4 | 1x65 |
vx3d-4x64 | 4 | 2 | 64 | 8 | 1x130 |
vx3d-8x128 | 8 | 4 | 128 | 16 | 1x260 |
vx3d-16x256 | 16 | 8 | 256 | 32 | 1x520 |
vx3d-24x384 | 24 | 12 | 384 | 48 | 1x780 |
vx3d-32x512 | 32 | 16 | 512 | 64 | 2x520 |
vx3d-48x768 | 48 | 24 | 768 | 96 | 2x780 |
vx3d-64x1024 | 64 | 32 | 1024 | 128 | 2x1024 |
vx3d-88x1408 | 88 | 44 | 1408 | 176 | 2x1430 |
vx3d-96x1536 | 96 | 48 | 1536 | 200 | 2x1560 |
vx3d-128x2048 | 128 | 64 | 2048 | 200 | 2x2080 |
vx3d-176x2816 | 176 | 88 | 2816 | 200 | 2x2860 |
Instance profile | vCPU | Cores | GiB RAM | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|---|
vx2d-2x28 | 2 | 1 | 28 | 4 | 1x60 |
vx2d-4x56 | 4 | 2 | 56 | 8 | 1x120 |
vx2d-8x112 | 8 | 4 | 112 | 16 | 1x240 |
vx2d-16x224 | 16 | 8 | 224 | 32 | 1x480 |
vx2d-44x616 | 44 | 22 | 616 | 80 | 1x1320 |
vx2d-88x1232 | 88 | 44 | 1232 | 80 | 2x1320 |
vx2d-144x2016 | 144 | 72 | 2016 | 80 | 2x2160 |
vx2d-176x2464 | 176 | 88 | 2464 | 80 | 2x2640 |
Ultra High Memory
Ultra High Memory profiles are hosted exclusively on the latest generation Intel® Xeon® Platinum Cascade Lake server hosts. This profile family offers our highest vCPU to memory ratio with 28 GiB of memory for every 1 vCPU of compute and up to 5.7 TiB of available RAM and is optimized for running memory intensive applications and in-memory database such as SAP HANA, Memcached, or Redis. All Very High Memory profiles are provisioned with temporary SSD-backed instance storage at no additional charge.
The following Ultra High Memory profiles are available for x86-64 processors:
Instance profile | vCPU / Cores | GiB RAM | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|
ux2d-2x56 | 2 / 1 | 56 | 4 | 1x60 |
ux2d-4x112 | 4 / 2 | 112 | 8 | 1x120 |
ux2d-8x224 | 8 / 4 | 224 | 16 | 1x240 |
ux2d-16x448 | 16 / 8 | 448 | 32 | 1x480 |
ux2d-36x1008 | 36 / 18 | 1008 | 64 | 1x1080 |
ux2d-48x1344 | 48 / 24 | 1344 | 80 | 2x720 |
ux2d-72x2016 | 72 / 36 | 2016 | 80 | 2x1080 |
ux2d-100x2800 | 100 / 50 | 2800 | 80 | 2x1500 |
ux2d-200x5600 | 200 / 100 | 5600 | 80 | 2x3000 |
GPU
The GPU profile family includes -v100
, -a100
, -l4
, and -l40S
and -h100
profiles. The GPU profile family includes profiles with and without instance storage.
- GPU
-v100
profiles include 1 or 2 NVIDIA V100 PCIe 16 GB GPUs. All OS images are supported on these GPU profiles. - Select availability GPU
-a100
profiles includes 8 NVIDIA A100 NVlink 80 GB GPUs. The a100 offering is available to select customers. This GPU profile supports only Linux OS images Ubuntu or RHEL. - GPU
-l4
profiles include NVIDIA L4 24 GB GPUs. - GPU
-l40S
profiles include NVIDIA L40S 48 GB GPUs. - Select availability GPU
-h100
profiles include NVIDIA H100 80 GB GPUs. The system is an HGX design. The H100 offering is available in the following regions and zones: London (eu-gb-2), Sydney (au-syd-2), Toronto (ca-tor-3), Madrid (eu-es-3), Washington DC (us-east-3), Tokyo (jp-tok-3), Sao Paulo (br-sao-1), Dallas (us-south-1), and Frankfurt (eu-de-2).
See Download drivers to review the most current versions that are supported. NVIDIA GPU drivers must be installed separately.
Instance profile | vCPU / Cores | GiB RAM | Type / Number of GPUs | Bandwidth Cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|---|
gx3-16x80x1l4 | 16 / 8 | 80 | l4 / 1 | 32 |
|
gx3-32x160x2l4 | 32 / 16 | 160 | l4 / 2 | 64 |
|
gx3-64x320x4l4 | 64 / 32 | 320 | l4 / 4 | 128 |
|
gx3-24x120x1l40s | 24 / 12 | 120 | l40s / 1 | 50 |
|
gx3-48x240x2l40s | 48 / 24 | 240 | l40s / 2 | 100 |
|
gx3d-160x1792x8h100 | 160 / 80 | 1792 | h100 / 8 | 200 | 8x7680 |
Instance profile | vCPU / Cores | GiB RAM | Type / Number of GPUs | Bandwidth Cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|---|
gx2-8x64x1v100 | 8 / 4 | 64 | v100 / 1 | 16 |
|
gx2-16x128x1v100 | 16 / 8 | 128 | v100 / 1 | 32 |
|
gx2-16x128x2v100 | 16 / 8 | 128 | v100 / 2 | 32 |
|
gx2-32x256x2v100 | 32 / 16 | 256 | v100 / 2 | 64 |
|
gx2-80x1280x8a100 | 80 / 40 | 1280 | a100 / 8 | 200 | 4x3200 |
Considerations for GPU profiles
When you create a -v100
, -a100
, -h100
, l4
, or l40S
GPU profile, keep the following recommendations in mind.
- During IBM Cloud periodic maintenance, GPU workloads aren't secure live migrated. Instead, the virtual server instance is restarted. You are notified 30 days in advance of any maintenance where the virtual server instance restarts. For more information, see Understanding cloud maintenance operations.
- If you are using GPU profiles, you need to install the NVIDA driver onto your virtual server instance. For more information, see Managing GPUs.
- If you are using GPU profiles, you might need to install the CUDA toolkit onto your virtual server instance. For more information, see Managing GPUs.
- For more information about persistent storage options, see Storage notes for profiles.
Storage Optimized
Storage Optimized profiles are hosted exclusively on Intel® Xeon® Platinum Cascade Lake servers. This profile family offers our highest vCPU to instance storage ratio with 300 GB of storage for every 1 vCPU and is optimized for running data lake and other workloads that require more intensive data capabilities. All storage optimized profiles are provisioned with temporary SSD-backed instance storage at no additional charge. For more information, see Lifecycle of instance storage.
Storage Optimized profiles use the Storage optimized (ox2) instance storage
quota, for instance, storage quota tracking. Unlike other profiles, which use the Instance storage
quota. For more information, see Quotas.
Storage Optimized profiles are available in the US South (Dallas), US East (Washington DC), United Kingdom (London), Germany (Frankfurt), EU Spain (Madrid), Japan (Tokyo), and Japan (Osaka) regions.
The following Storage Optimized profiles are available for x86-64 processors:
Instance profile | vCPU | Cores | GiB RAM | Bandwidth cap (Gbps) | Instance storage (GB) | Interface type |
---|---|---|---|---|---|---|
ox2-2x16 | 2 | 1 | 16 | 4 | 1x600 | virtio_blk |
ox2-4x32 | 4 | 2 | 32 | 8 | 1x1200 | virtio_blk |
ox2-8x64 | 8 | 4 | 64 | 16 | 2x1200 | virtio_blk |
ox2-16x128 | 16 | 8 | 128 | 32 | 2x2400 | virtio_blk |
ox2-32x256 | 32 | 16 | 256 | 64 | 3x3200 | virtio_blk |
ox2-64x512 | 64 | 32 | 512 | 80 | 6x3200 | virtio_blk |
ox2-96x768 | 96 | 48 | 768 | 80 | 9x3200 | virtio_blk |
ox2-128x1024 | 128 | 64 | 1024 | 80 | 12x3200 | virtio_blk |
Confidential computing profiles
Select availability
The following profiles support Confidential computing with Intel Software Guard Extensions (SGX), Confidential computing with Intel Trusted Domain Extension (TDX), and secure boot.
For more information about confidential computing, see Confidential computing for x86 Virtual Servers for VPC.
Instance profile | vCPU / Cores | GiB RAM | EPC (SGX) capacity (GiB) | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|---|
bx3dc-2x10 | 2 / 1 | 10 | 4 | 4 | 1x65 |
bx3dc-4x20 | 4 / 2 | 20 | 8 | 8 | 1x130 |
bx3dc-8x40 | 8 / 4 | 40 | 16 | 16 | 1x260 |
bx3dc-16x80 | 16 / 8 | 80 | 32 | 32 | 1x520 |
bx3dc-24x120 | 24 / 12 | 120 | 48 | 48 | 1x780 |
bx3dc-32x160 | 32 / 16 | 160 | 64 | 64 | 2x520 |
bx3dc-48x240 | 48 / 24 | 240 | 96 | 96 | 2x780 |
bx3dc-64x320 | 64 / 32 | 320 | 128 | 128 | 2x1024 |
bx3dc-96x480 | 96 / 48 | 480 | 192 | 480 | 2x1560 |
Instance profile | vCPU / Cores | GiB RAM | EPC (SGX) capacity (GiB) | Bandwidth cap (Gbps) | Instance storage (GB) |
---|---|---|---|---|---|
cx3dc-2x5 | 2 / 1 | 5 | 2 | 4 | 1x65 |
cx3dc-4x10 | 4 / 2 | 10 | 4 | 8 | 1x130 |
cx3dc-8x20 | 8 / 4 | 20 | 8 | 16 | 1x260 |
cx3dc-16x40 | 16 / 8 | 40 | 16 | 32 | 1x520 |
cx3dc-24x60 | 24 / 12 | 60 | 24 | 48 | 1x780 |
cx3dc-32x80 | 32 / 16 | 80 | 32 | 64 | 2x520 |
cx3dc-48x120 | 48 / 24 | 120 | 48 | 96 | 2x780 |
cx3dc-64x160 | 64 / 32 | 160 | 64 | 128 | 2x1024 |
cx3dc-96x240 | 96 / 48 | 240 | 96 | 192 | 2x1560 |
cx3dc-128x320 | 128 / 64 | 320 | 128 | 200 | 2x2860 |
Bandwidth allocation
Instance bandwidth is allocated between volume bandwidth and networking bandwidth. The bandwidth capacity (Bandwidth Cap) is determined by the virtual server profile that you select during instance provisioning. For example, a bx2-2x8 balanced server profile allows a bandwidth cap of 4 Gbps.
Bandwidth allocation between storage and networking
The initial volume and network bandwidth allocation depends on the bandwidth that is set by the instance profile that you selected. You can also see the bandwidth allocations in the profile information during instance creation in the console. The bandwidth allocation between Storage and Network can be changed on the instance details page after you provision an instance.
For example, for the bx2-2x8 profile, you might have the following bandwidth configuration:
- Storage: 1 Gbps
- Network: 3 Gbps
You can adjust the amount of overall bandwidth that is provided to storage volumes within the overall instance limits. However, both volume and network bandwidth must be at least 500 MBps each. For example, to allow more bandwidth for volumes, you can apportion the bx2-2x8 example in equal allocations:
- Storage: 2 Gbps
- Network: 2 Gbps
For more information, see Bandwidth allocation for instance profiles and Adjusting bandwidth allocation by using the UI.
For more information, see Bandwidth allocation for instance profiles and Adjusting bandwidth allocation by using the CLI.
For more information, see Bandwidth allocation for instance profiles and Adjusting total storage bandwidth allocation from the API.
Bandwidth allocation with multiple data volumes
The bandwidth for volumes is divided between all the attached volumes. To help ensure reasonable boot times, a minimum of 393 MBps is allocated to the primary boot volume. The remaining bandwidth is divided between the data volumes. You can attach up to 12 data volumes to your instance. The data volumes are assigned bandwidth that is proportional to their maximum bandwidth. For example, if you have an instance with four identical data volumes, the overall volumes bandwidth is divided equally among them. If you are using only one volume at a time, then that volume still gets only the bandwidth that is assigned to it, one-fourth of the overall volumes bandwidth. For more information, see Bandwidth allocation for Block Storage volumes.
Bandwidth allocation with multiple network interfaces
You can add up to 15 network interfaces for your virtual server instance, depending on the vCPU count that is included in the instance profile.
- 2-16 vCPUs: Up to five network interfaces
- 17-48 vCPUs: Up to 10 network interfaces
- 49 or more vCPUs: Up to 15 network interfaces
With multiple network interfaces, bandwidth is distributed evenly across the network interfaces that are attached to the virtual server instance.
For more information, see Managing network interfaces.
Block Storage volume notes for profiles
When you create secondary data volumes, you select a volume profile that best meets your requirements. Volume profiles are available as three predefined tiers or as a custom profile. These volume profiles relate to virtual server instance profiles:
- A 3 IOPS general-purpose tier profile provides IOPS/GB performance that is suitable for a virtual server instance Balanced profile.
- A 5-IOPS tier profile provides IOPS/GB performance that is suitable for a virtual server instance Compute profile.
- A 10-IOPS tier profile provides IOPS/GB performance that is suitable for a virtual server instance Memory profile.
Viewing profile configurations
You can view available profile configurations by using the IBM Cloud console or the CLI. In the IBM Cloud console, you can select from popular profile configurations that support the most common use cases.
Understanding the naming rule of the profiles
The following information describes the profile naming rule.
The first character represents the profile families. Different profile families have different ratios of vCPU to memory and other characteristics that are designed for different workloads.
- "b": balanced family of profiles
- "c": compute family of profiles (higher on the CPUs)
- "m": memory family of profiles (higher on the memory)
- "u": ultra high memory family of profiles, 1 vCPU to 28 GiB of memory ratio
- "v": very high memory family of profiles, 1 vCPU to 14 GiB of memory ratio
- "g": GPU profiles, which is a 1:8 or 1:16 ratio
- "o": storage optimized family of profiles, 1 vCPU to 8 GiB memory ratio and 1 vCPU to 300 GB instance storage ratio
The second character represents the CPU architecture.
- "x": x86_64
- "z": s390x
The third character represents the generation of the IBM Cloud infrastructure where the profile is provisioned.
If the fourth character is a "d", such as bx2d, then a defined quantity of instance storage is provisioned with the virtual server.
The characters after "-" represents the number of vCPUs and the size of RAM (GiB). For example, "2x8" means that this profile has 2 vCPU and 8 GiB of RAM.
Using “bx2-4x16” as an example, you can know from the name that it is a balanced profile that provides 4 vCPUs of compute and 16 GiB of memory. The profile is deployed on an x86-based host and is for the second-generation VPC.
Viewing instance profiles in the UI
- In the IBM Cloud console, go to Navigation Menu icon > Infrastructure > Compute > Virtual server instances.
- From the Virtual server instances page, click New instance.
- You can either select a profile configuration from Popular profiles or click All profiles to view more configurations.
Viewing instance profiles from the CLI
To view the list of available instance profiles by using the CLI, run the following command:
ibmcloud is instance-profiles
Viewing instance profiles with the API
To view the list of available instance profiles by using the API, you can call the List all instance profiles API.
The following request example lists the available instance profiles. When you call the API, replace the API endpoint and IAM token with the values from your enterprise. For more information about the $vpc_api_endpoint
and $iam_token
variables, see the Authentication and Endpoint URLs sections in Virtual Private Cloud API Introduction.
curl -X GET \
"$vpc_api_endpoint/v1/instance/profiles?version=2021-02-23&generation=2" \
-H "Authorization: $iam_token"
Next generation instance profiles
The 3rd generation of IBM Cloud Virtual Servers for VPC are available to provision in all regions. This new generation features virtual server profile families that are hosted exclusively on Intel 4th Generation Xeon Scalable processors to provide the most powerful and performant general-purpose profiles available. These 3rd generation profiles provide the following enhancements:
- Improved performance with DDR 5 memory DIMMs, PCI Gen 5 interconnects, and more memory per vCPU than prior generation profiles.
- A wide variety of profiles sizes with core to memory ratios optimized to maximize performance and economics for intensive workloads.
- Enhanced integrated accelerators that feature AMX-512, AVX, and enhanced crypto acceleration.
- Instances are started by default with Open Virtual Machine Format (OVMF), and run in Unified Extensible Firmware Interface (UEFI) mode for enhanced security.
- Local instance storage is included with all profiles for ease of access to temporary storage and swap space. For more information about the temporary nature of instance storage, see Lifecycle of instance storage.
- A 3rd generation profile can be resized to a 2nd generation profile. A 2nd generation profile can be resized to a 3rd generation profile. For more information, see Resizing between Gen 2 and Gen 3 profiles.
Intel Hyper-Threading Technology
All Intel® x86-64 servers have Hyper-Threading enabled by default. Intel® Hyper-Threading Technology is a term that describes simultaneous multithreading (SMT). Hyper-Threading Technology splits each physical core into two virtual processors. Hyper-Threading Technology is like taking a wide road with a single lane and making it into two relatively narrower lanes. The two-lane highway provides better service over the single-lane road if traffic is moving slow and fast. Hyper-Threading Technology provides better application performance for File I/O, Network I/O, and other slower operations mixed with CPU-intensive operations. The performance advantage of Hyper-Threading Technology typically ranges in the range 0 - 30% over a single-thread mode. Some applications might also see a drop in performance.
If you want to disable Intel® Hyper-Threading, see Disabling Intel Hyper-Threading Technology.
Next steps
After you choose a profile, you're ready to create an instance.