machine shapes.md

Machine Shapes

Namespace provides flexible compute resources through a comprehensive selection of machine shapes designed to meet diverse workload requirements. Machine shapes define the CPU, memory, and storage resources allocated to your compute instances, allowing you to optimize performance and cost for your specific use cases.

Understanding Machine Shapes

Machine shapes in Namespace follow a simple AxB notation format, where A represents the number of virtual CPUs (vCPUs) and B represents the amount of memory in gigabytes (GB). For example, a 4x16 machine shape provides 4 vCPUs and 16 GB of RAM.

This standardized format makes it easy to identify and select the appropriate resources for your workload, whether you're running lightweight development tasks or memory-intensive applications.

Available Configurations

Each platform supports its own shape configurations, and the following tables list the most common shapes for each. All prices are per minute.

Linux

Shape vCPU RAM Disk size Multiplier Prepaid Overage
1x2 Minimal 1 2 GB 40 GB 1 $0.001 $0.0015
2x4 Optimal 2 4 GB 48 GB 2 $0.002 $0.003
2x8 2 8 GB 64 GB 4 $0.004 $0.006
4x8 Optimal 4 8 GB 64 GB 4 $0.004 $0.006
4x16 4 16 GB 96 GB 8 $0.008 $0.012
8x16 Optimal 8 16 GB 96 GB 8 $0.008 $0.012
8x32 8 32 GB 160 GB 16 $0.016 $0.024
16x32 Optimal 16 32 GB 160 GB 16 $0.016 $0.024
16x64 16 64 GB 288 GB 32 $0.032 $0.048
32x64 Optimal 32 64 GB 288 GB 32 $0.032 $0.048
32x128 32 128 GB 544 GB 64 $0.064 $0.096
32x256 32 256 GB 1056 GB 128 $0.128 $0.192
32x512 32 512 GB 2080 GB 256 $0.256 $0.384
64x128 64 128 GB 544 GB 80 $0.08 $0.12
64x256 64 256 GB 1056 GB 160 $0.16 $0.24
64x512 64 512 GB 2080 GB 320 $0.32 $0.48

Windows

Shape vCPU RAM Disk size Multiplier Prepaid Overage
1x2 Minimal 1 2 GB 40 GB 2 $0.002 $0.003
2x4 Optimal 2 4 GB 48 GB 4 $0.004 $0.006
2x8 2 8 GB 64 GB 8 $0.008 $0.012
4x8 Optimal 4 8 GB 64 GB 8 $0.008 $0.012
4x16 4 16 GB 96 GB 16 $0.016 $0.024
8x16 Optimal 8 16 GB 96 GB 16 $0.016 $0.024
8x32 8 32 GB 160 GB 32 $0.032 $0.048
16x32 Optimal 16 32 GB 160 GB 32 $0.032 $0.048
16x64 16 64 GB 288 GB 64 $0.064 $0.096
32x64 Optimal 32 64 GB 288 GB 64 $0.064 $0.096
32x128 32 128 GB 544 GB 128 $0.128 $0.192
32x256 32 256 GB 1056 GB 256 $0.256 $0.384
32x512 32 512 GB 2080 GB 512 $0.512 $0.768

macOS

Shape vCPU RAM Disk size Multiplier Prepaid Overage
4x7 4 7 GB 80 GB 40 $0.04 $0.06
6x14 6 14 GB 104 GB 60 $0.06 $0.09
6x28 6 28 GB 160 GB 75 $0.075 $0.1125
8x14 8 14 GB 112 GB 80 $0.08 $0.12
12x28 12 28 GB 160 GB 120 $0.12 $0.18
12x56 12 56 GB 272 GB 180 $0.18 $0.27
16x28 16 28 GB 176 GB 160 $0.16 $0.24
16x56 16 56 GB 272 GB 200 $0.20 $0.30

Linux on Apple Silicon

Shape vCPU RAM Disk size Multiplier Prepaid Overage
2x4 2 4 GB 64 GB 14 $0.014 $0.021
2x8 2 8 GB 80 GB 28 $0.028 $0.042
4x8 4 8 GB 80 GB 28 $0.028 $0.042
4x12 4 12 GB 96 GB 42 $0.042 $0.063
4x16 4 16 GB 112 GB 56 $0.056 $0.084
6x12 6 12 GB 96 GB 42 $0.042 $0.063
6x14 6 14 GB 104 GB 49 $0.049 $0.0735
8x16 8 16 GB 112 GB 56 $0.056 $0.084
8x32 8 32 GB 176 GB 112 $0.112 $0.168
12x16 12 16 GB 144 GB 84 $0.084 $0.126
12x24 12 24 GB 144 GB 84 $0.084 $0.126
12x32 12 32 GB 176 GB 112 $0.112 $0.168
12x56 12 56 GB 272 GB 196 $0.196 $0.294
16x32 16 32 GB 176 GB 112 $0.112 $0.168
16x56 16 56 GB 272 GB 196 $0.196 $0.294

Storage Allocation

Namespace automatically provisions ephemeral storage based on your selected machine shape. The storage allocation scales dynamically to ensure adequate disk space for your workload.

The disk space for an instance is calculated using the following formula:

Linux & Windows

$$ disk_size = 32GB + \max(vCPU_count, RAM_GB / 2) \times 8GB $$

macOS and Linux on Apple Silicon

$$ disk_size = 48GB + \max(vCPU_count, RAM_GB / 2) \times 8GB $$

Non-Standard shapes

Namespace allows you to freely choose your shape configuration to best match your workflow requirements. For example, you can grow your memory allocation to 64GB while keeping the vCPU count at 4, forming a 4x64 shape.

Standard shapes for Linux and Windows are multiples of 1 vCPU and 2GB RAM. For example: 1x2 (minimum), 2x4 (tiny), 4x8 (small), 8x16 (medium), 16x32 (large), 32x64 (maximum).

Non-standard shapes consume the per-minute cost of the smallest standard shape that contains them. When billing an instance, Namespace calculates how many 1 vCPU x 2GB RAM units were consumed for that instance:

$$ unit_count = \max(vCPU_count, RAM_GB / 2) $$

Contact support@namespace.so if you have special shape requirements.

High-Memory and High-CPU Instances

For workloads requiring exceptional memory capacity like data processing and analytics pipelines, Namespace offers extended memory configurations:

Note: High-memory and high-CPU instances require special access and may not be available on all subscription plans. Contact support@namespace.so to request access for your workspace.

High-CPU (64 vCPU) instances have an additional 1.25x unit multiplier.

Billing and Cost Optimization

Machine shapes are billed on a per-minute basis with a one-minute minimum charge. For more details on the usage and billing calculation, check out Billing and Limits →.

When trying to optimize performance and cost, picking the right shape is an important step:

For assistance with machine shape selection, high-memory or high-CPU instance access, or performance optimization, reach out to our support team.

Unit minute accounting

We account for a minimum of 1 minute for each instance, but round the next 15 seconds down. For example, 30 seconds of usage is 1 billable minute, 70 seconds is 1 billable minute, and 150 seconds are billed as 3 minutes. Jobs that run on Windows or macOS consume minutes at a higher rate than jobs on Linux. Unit minutes usage per shape and platform can be found above.

An instance is only accounted for the time it runs a job. Instances that never pick up a job, such as runners whose job was cancelled or picked up by another matching runner, consume no unit minutes.