Flavors
Understanding Flavors
Flavors define the resource specifications of a virtual machine, including CPU, memory, disk space, and additional features.
OpenStack VM flavors come in various configurations, each tailored to meet specific workload requirements. Key parameters of a flavor include:
- CPU: Specifies the number of virtual CPUs (vCPUs) allocated to the VM.
- RAM: Determines the amount of memory (in MB or GB) assigned to the VM.
- Disk: Indicates the size of the VM's ephemeral disk (Not used in Switch Cloud Compute)
Important
Switch has defined the flavors available for Switch Cloud Compute.
Switch Cloud Compute Flavors Naming
Pattern:
c<CPU>r<RAM>, where:
CPU: Number of vCPUs (3 digits)RAM: RAM in GB (3 digits)
Examples:
c001r001: 1 vCPU, 1GB RAMc004r016: 4 vCPUs, 16GB RAM
Purposes of Different Flavors
General-Purpose Flavors:
These flavors offer a balanced combination of CPU, and memory, suitable for a wide range of workloads, including web servers, and development environments.
| Name | vCPUs | RAM |
|---|---|---|
| c001r002 | 1 | 2 GB |
| c001r004 | 1 | 4 GB |
| c002r004 | 2 | 4 GB |
| c002r008 | 2 | 8 GB |
| c004r008 | 4 | 8 GB |
| c004r016 | 4 | 16 GB |
| c008r016 | 8 | 16 GB |
| c008r032 | 8 | 32 GB |
| c016r032 | 16 | 32 GB |
| c016r064 | 16 | 64 GB |
| c032r064 | 32 | 64 GB |
| c032r128 | 32 | 128 GB |
Memory-Optimized Flavors:
Designed for memory-intensive applications such as in-memory databases, caching servers, and big data analytics. These flavors allocate a significant amount of RAM to support data-intensive operations efficiently.
| Name | vCPUs | RAM |
|---|---|---|
| c001r008 | 1 | 8 GB |
| c002r016 | 2 | 16 GB |
| c004r032 | 4 | 32 GB |
| c008r064 | 8 | 64 GB |
| c016r128 | 16 | 128 GB |
| c032r256 | 32 | 256 GB |
Compute-Optimized Flavors:
Ideal for CPU-bound workloads that require high computational power, such as video encoding, rendering, and scientific simulations. These flavors prioritize CPU resources to optimize performance.
| Name | vCPUs | RAM |
|---|---|---|
| c001r001 | 1 | 1 GB |
| c002r002 | 2 | 2 GB |
| c004r004 | 4 | 4 GB |
| c008r008 | 8 | 8 GB |
| c016r016 | 16 | 16 GB |
| c032r032 | 32 | 32 GB |
| c032r064 | 32 | 64 GB |
GPU Flavors:
For GPU workloads.
| Name | vCPUs | RAM | Ephemeral | GPU type |
|---|---|---|---|---|
| gpu-1xL40S.d.c040r345 | 40 | 345 GiB | N/A | 1x NVIDIA L40S |
| gpu-2xL40S.d.c080r690 | 80 | 690 GiB | N/A | 2x NVIDIA L40S |
| gpu-2xH100NVL.d.c040r500 | 40 | 500 GiB | N/A | 2x NVIDIA H100 NVL |
| gpu-1xH200.d.e7000.c028r345 | 28 | 345 GiB | 7000 GiB | 1x NVIDIA H200 |
| gpu-2xH200.d.e14000.c056r690 | 56 | 690 GiB | 14000 GiB | 2x NVIDIA H200 |
| gpu-4xH200.d.e14000.c112r1380 | 112 | 1380 GiB | 14000 GiB | 4x NVIDIA H200 |
Tip
GPU flavors require a high amount of resources: please have your project quotas adjusted beforehand.
Warning
A GPU instance needs to map the address space of the available GPU memory into its address space. Because of the address space size, booting a VM may take several minutes.
Warning
Ephemeral disk space is high performance local disk storage, similar to commonly referred scratch space. During a shelve/unshelve or resize-operation the data on those disks will be lost. This storage is NOT using redundant hardware, a hardware failure might result in data loss. For persistent and redundant storage, please use volumes or S3 storage.
Use Cases for Different Flavors
By selecting the appropriate flavor for each application, organizations can optimize performance, enhance resource utilization, and deliver exceptional user experiences across a wide range of use cases.
- Web Application Hosting: General-purpose flavors are well-suited for hosting web applications, providing sufficient CPU and memory resources to handle incoming traffic efficiently.
- Big Data Processing: Memory-optimized flavors excel in processing large datasets and running memory-intensive analytics tasks, making them ideal for big data processing frameworks like Apache Spark and Hadoop.
- Performance Computing: Compute-optimized flavors are indispensable for running computationally intensive simulations, mathematical modeling, and scientific research.