DataSphere pricing
What goes into the cost of using DataSphere
When using DataSphere, you pay for the use of computing resources: the computation and instance running time is charged per second.
Pricing unit
A pricing unit means a single billing unit. The number of billing units spent on calculations depends on:
-
Capacity of the computing resources being used.
-
Time spent on calculations.
The calculation time is rounded up to an integer number of seconds.
One billing unit equals the cost of using one CPU core for one second. The number of units depends on the computing resource configuration.
DataSphere Notebook
When performing computations and training models in DataSphere notebooks, you pay for the whole time the VM is used since the first computations are run in a notebook until the VM is released, regardless of whether any computations were performed after the first run. If you run computations using Data Proc clusters, they are charged separately.
You are separately charged for data storage in datasets and project storage expansion.
Running jobs in DataSphere Jobs
When you run computations remotely using DataSphere Jobs, you pay for the computation time in the selected computing resource configuration. You also pay for storing data required to run jobs: their cache, logs, input data, and execution results.
DataSphere Inference
When deploying models with DataSphere Inference, you pay for the uptime of each node instance: from its start to deletion.
Using Data Proc clusters
The cost of computations in Data Proc clusters includes:
- Computing resources of the selected DataSphere configuration.
- The entire lifetime of the Data Proc cluster according to the Yandex Data Proc pricing policy.
Learn more about integration with Data Proc.
Disk space usage
Each DataSphere project has a free-of-charge storage amount. If you increase the project size quota
Data storage inside datasets is charged separately.
If you need to store large amounts of data, you can use Yandex Object Storage. In this case, data storage will be charged according to the Object Storage pricing policy.
Model storage is charged separately. Its pricing depends on the model size that is calculated in GB and rounded up to an integer.
In all calculations, 1 GB = 230 bytes.
The monthly usage rate is based on 720 hours per month.
Examples of cost calculation
DataSphere Notebook
Cost of using DataSphere with the following parameters:
- Computing resources: g1.1 configuration with 8 CPUs and 1 GPU.
- VM usage time: 10 minutes.
The cost is calculated as follows:
72 × 600 = 43,200 units for VM usage
43,200 × 0.0000096 = $0.4147200Total: $0.4147200 is the cost of using DataSphere.
Where:
- 72: Number of units per g1.1 configuration.
- 600: VM usage time in seconds.
- $0.0000096: Cost per unit.
DataSphere Inference
Cost of using DataSphere Inference with the following parameters:
- Number of instances per node: 2.
- Instance configuration: g1.1 with 8 CPUs and 1 GPU.
- Node runtime: 2 days 3 hours 36 minutes.
Converting node runtime into seconds:
2 × (24 × 60 × 60) + 3 × (60 × 60) + 36 × 60 =172,800 + 10,800 + 2,160= 185,760
Where:
- (24 × 60 × 60): Number of seconds in 24 hours.
- (60 × 60): Number of seconds in 1 hour.
- 60: Number of seconds in 1 minute.
The cost is calculated as follows:
72 × 185,760 × 2 = 26,749,440 units for running a node
26,749,440 × 0.0000096 = $256.7946240Total cost of using DataSphere: $256.7946240.
Where:
- 72: Number of units per g1.1 configuration.
- 185,760 : Node runtime in seconds.
- 2: Number of instances per node.
- $0.0000096: Cost per unit.
Pricing
The prices of the configurations are valid for computations in DataSphere Notebook, DataSphere Jobs jobs, and the use of DataSphere Inference.
Unit cost
Number | Cost per 1 second of computations, without VAT |
---|---|
One unit | $0.0000096 |
Computing resource
Configuration | Number of units per configuration |
Cost per 1 second of computations, without VAT |
---|---|---|
c1.4 (4 vCPUs, 0 GPU) | 4 | $0.000038 |
c1.8 (8 vCPUs, 0 GPU) | 8 | $0.000077 |
c1.32 (32 vCPUs, 0 GPU) | 32 | $0.000308 |
c1.80 (80 vCPUs, 0 GPU) | 80 | $0.000769 |
g1.1 (8 vCPUs, 1 GPU V100) | 72 | $0.000692 |
g1.2 (16 vCPUs, 2 GPUs V100) | 144 | $0.001385 |
g1.4 (32 vCPUs, 4 GPUs V100) | 288 | $0.002769 |
g2.1 (28 vCPUs, 1 GPU A100) | 116 | $0.001115 |
g2.2 (56 vCPUs, 2 GPUs A100) | 232 | $0.00223 |
g2.4 (112 vCPUs, 4 GPUs A100) | 464 | $0.004462 |
g2.8 (224 vCPUs, 8 GPUs A100) | 928 | $0.008923 |
gt4.1 (4 vCPUs, 1 GPU T4) | 36 | $0.0003456 |
Storing data inside DataSphere
Resource | Cost of 1 GB per month, without VAT |
---|---|
Project storage size, up to 10 GB | Free of charge |
Project storage size, over 10 GB | $0.095280 |
Datasets | $0.095437 |
Model | $0.095437 |
DataSphere Jobs data | Free of charge |
Egress traffic
When using the service, you pay for traffic from Yandex Cloud to the internet. Traffic between Yandex Cloud services and incoming internet traffic are free of charge.
The first 100 GB of outgoing traffic are provided free of charge every month.
The minimum billing unit is 1 MB.
Resource category | Cost of 1 GB |
---|---|
Outgoing traffic via NAT gateway, first 10 GB per month | Free |
Outgoing traffic via NAT gateway, over 10 GB per month | $0.012240 |