The billing of Data Lake Compute (DLC) consists of two parts: computing resources and storage resources. Computing resources are categorized into private and public engines, and storage resources are charged only when the managed storage service of DLC is used.
Computing Resource Billing
Billing Modes
DLC supports pay-as-you-go and yearly/monthly subscription billing modes, as detailed in the table below. For details about pay-as-you-go billing, see Postpaid Billing. For details about yearly/monthly subscription billing, see Prepaid Billing. Note:
1. The public engine is deprecated and only available for existing users.
2. The computing capability of 1 compute unit (CU) is approximately equal to 1 CPU core and 4 GB memory. You can estimate the number of required CUs based on your actual business conditions or submit a ticket for assistance. |
Public Engine *Deprecated | Pay-as-you-go | Requires no configuration management. Supports billing by scanned data volume. No fees are incurred if resources are idle. | Works well in temporary computing scenarios with a small amount of data. |
Private Engine | Pay-as-you-go | Supports resource scale configurations. Resources are dedicated and can be scaled as needed. Supports billing by CU. Clusters can be suspended when there is no task, which incurs no fees. | Works well in irregular computing scenarios with a medium amount of data. |
| Monthly subscription | Supports resource scale configurations. Resources are dedicated and can be scaled as needed. Supports billing by CU. Yearly/Monthly subscription clusters are available at any time with no waiting. Elastic resources support pay-as-you-go. | Works well in computing scenarios with a large and stable amount of data. |
Pricing
Public Engine
The billing rules are as follows:
A public engine is billed by scanned data volume in successful query tasks, with a bill generated every hour. You can view specific bills in the Billing Center. If less than 34 MB of data is scanned for a single SQL task, it will be billed as 34 MB (about USD 0.00015).
Fees will be incurred if the query task succeeds or the task is manually canceled but data is scanned. No fees will be incurred when resources are being scheduled or if the query task fails.
Public engine fee = Scanned data volume x Unit price
|
Billed by Scanned Data Volume | 0.0045 |
When a public engine is used, since DLC adopts serverless architecture, computing resources need to be scheduled for task execution for the first time over a period of time. This may take a longer time.
Private Engine
Pay-As-You-Go
A pay-as-you-go data engine is scalable. Billing is based on the actual usage duration of a cluster in seconds. When you purchase it, the system will freeze the fee for 1-hour usage of the minimum cluster specifications. A bill will be generated every hour. You can view specific bills in the Billing Center. Pay-as-you-go private engine fee = Used CUs per hour x Unit price
Note
1. In pay-as-you-go billing mode, an engine will keep running at the purchased minimum cluster count. You can suspend the clusters when you do not want to use them to avoid fees.
2. To purchase a data engine with more than 256 CUs, submit a ticket to contact us. We will reserve resources for you. Monthly Subscription
In monthly subscription billing mode, an engine will keep running at the purchased minimum cluster count. If scale-out is triggered, added cluster resources will adopt the pay-as-you-go billing mode. You can view specific bills in the Billing Center. A monthly subscription engine does not incur pay-as-you-go fees if it is not scaled out.
Monthly subscription fee = Cluster specifications x Minimum cluster count x Number of purchase months x Unit price of monthly subscription billing
Monthly subscription scale-out fee = Cluster specifications x Number of clusters added by scale-out x Unit price of pay-as-you-go billing
Example: Customer A purchases a monthly subscription data engine, with cluster specifications of 16 CUs, a minimum cluster count of 2 and a maximum cluster count of 5 for 1 month, the monthly subscription fee is calculated as follows: 16 x 2 x 1 x 22 = USD 704. The private data engine will run with two 16-CU clusters for 1 month without incurring additional fees. If the engine is scaled out to 5 clusters and runs for 1 hour, the three additional 16-CU clusters will incur the following pay-as-you-go fee: 16 x (5 – 2) x 1 x 0.05 = USD 2.4.
1. CU unit price
Description
Addition to the allowlist is required in a few regions. The available regions are shown on the purchase page. For any requirement, submit a ticket for consultation. The prices of DLC CUs in different regions are as follows:
|
Hong Kong (China) | 22 | 0.05 |
Singapore | 22 | 0.05 |
Jakarta | 22 | 0.05 |
Bangkok | 22 | 0.05 |
Seoul | 28 | 0.09 |
Virginia | 22 | 0.05 |
Tokyo | 28 | 0.09 |
Silicon Valley | 28 | 0.09 |
Frankfurt | 22 | 0.05 |
2. GPU unit price
To support scenarios such as deep learning algorithm training, DLC has added the GPU computing capability in the private engine. Currently, GPU computing engines can only be purchased with the yearly/monthly subscription mode. When purchasing, users select and purchase a machine based on its model and specifications.
Monthly subscription fee = Unit price x Number of machines x Number of purchase months
Example: Customer A purchases a monthly subscription GPU data engine with the model of GN10Xp, specifications of 10 CUs/1 GPU, and 1 machine for 1 month, the monthly subscription fee is calculated as follows: 1726 x 1 x 1 = USD 1726. The private data engine will run a 10 CUs/1 GPU GN10Xp cluster for 1 month without incurring additional fees. If the configurations are changed to GN10Xp model with 20 CUs/2 GPUs and 1 machine for 1 month, the monthly subscription fee is as follows: 3352 x 1 x 1 = USD 3452.
Note:
Currently, GPU and related features are allowlist features. For any requirement, you can submit a ticket for consultation. |
GN10Xp V100-32G | Hong Kong (China) and Silicon Valley | 10CU/1GPU | 1726 |
|
| 20CU/2GPU | 3452 |
|
| 40CU/4GPU | 6903 |
|
| 80CU/8GPU | 13806 |
| Frankfurt | 10CU/1GPU | 1812 |
|
| 20CU/2GPU | 3624 |
|
| 40CU/4GPU | 7248 |
|
| 80CU/8GPU | 14495 |
| Singapore, Bangkok, and Tokyo | 10CU/1GPU | 1760 |
|
| 20CU/2GPU | 3519 |
|
| 40CU/4GPU | 7038 |
|
| 80CU/8GPU | 14076 |
| Seoul | 10CU/1GPU | 1661 |
|
| 20CU/2GPU | 3322 |
|
| 40CU/4GPU | 6644 |
|
| 80CU/8GPU | 13287 |
GN7 T4-16G | Hong Kong (China), Singapore, Jakarta, Tokyo, and Silicon Valley | 8CU/1GPU | 616 |
|
| 20CU/1GPU | 983 |
|
| 32CU/1GPU | 1351 |
|
| 40CU/2GPU | 1966 |
|
| 80CU/4GPU | 3932 |
| Bangkok and Frankfurt | 8CU/1GPU | 674 |
|
| 20CU/1GPU | 1077 |
|
| 32CU/1GPU | 1481 |
|
| 40CU/2GPU | 2154 |
|
| 80CU/4GPU | 4307 |
| Virginia | 8CU/1GPU | 463 |
|
| 20CU/1GPU | 727 |
|
| 32CU/1GPU | 991 |
|
| 40CU/2GPU | 1453 |
|
| 80CU/4GPU | 2905 |
| Seoul | 8CU/1GPU | 532 |
|
| 20CU/1GPU | 849 |
|
| 32CU/1GPU | 1166 |
|
| 40CU/2GPU | 1698 |
|
| 80CU/4GPU | 3395 |
PNV5b L20-48G | Hong Kong (China) | 48CU/1GPU | 1691 |
|
| 96CU/2GPU | 3381 |
|
| 192CU/4GPU | 6762 |
|
| 384CU/8GPU | 13523 |
| Jakarta and Tokyo | 48CU/1GPU | 1864 |
|
| 96CU/2GPU | 3727 |
|
| 192CU/4GPU | 7453 |
|
| 384CU/8GPU | 14906 |
| Singapore and Frankfurt | 48CU/1GPU | 2085 |
|
| 96CU/2GPU | 4169 |
|
| 192CU/4GPU | 8337 |
|
| 384CU/8GPU | 16674 |
Storage Resource Billing
DLC provides the data storage capability. After you enable the native table feature, the native table data is stored in the managed storage of DLC.
The DLC managed storage is used to store data such as native table data, adaptive shuffle data, query result data, and program packages.
Billing Modes
Managed storage fees of DLC are charged in pay-as-you-go billing mode, which consists of storage usage fee and request fee, as detailed below.
|
Storage Usage Fee | Calculated based on the storage usage at a unit price that varies by storage type. | Daily storage usage fee = Monthly storage usage unit price/30 x Daily storage usage Daily storage usage = Sum of "5-minute storage usage" for the day/288 (number of statistical points) |
Request Fee | Calculated based on the number of requests at a unit price that varies by storage type. | Request fee = Unit price per 10,000 requests x Daily cumulative request count/10,000 |
Pricing
|
Hong Kong (China) | 0.022 | 0.002 |
Singapore | 0.017 | 0.003 |
Jakarta | 0.022 | 0.003 |
Bangkok | 0.017 | 0.002 |
Seoul | 0.024 | 0.002 |
Virginia | 0.02 | 0.002 |
Tokyo | 0.017 | 0.002 |
Silicon Valley | 0.024 | 0.002 |
Frankfurt | 0.02 | 0.002 |
Example:
On November 1, 2020, customer A uploaded 10 GB of standard data to the DLC managed bucket in Hong Kong (China), generating 100 requests. As fees for a day are settled the next day, then:
Standard storage usage fee: settled daily starting from November 2, 2020.
Request fee: settled on November 2, 2020.
An analysis is performed as follows based on the two billing methods:
Standard storage usage fee = 0.022 USD/GB/month x 10 GB = USD 0.22/month
Request fee = 0.002 USD/10,000 requests x 100 requests/10,000 = USD 0.00002
In summary, the total bill for customer A in November is calculated as follows: USD 0.22 + USD 0.00002 = USD 0.22002.