topic Online Feature Table : Storage in Machine Learning

Online Feature Table : Storage

AnkurMittal008 — Tue, 17 Dec 2024 07:24:28 GMT

Databricks Online Feature Table is in public preview , And we have some questions on this

1) What storage is being used for Online Feature table's Data. Our offline feature table is stored in Unity Catalog managed S3 bucket (Customer AWS ). Does online feature table use databrick's internal storage or customer's Cloud ?

2) https://www.databricks.com/product/pricing/online-tables How we can get the storage cost information. For capacity units we can use billing_origin_product = ONLINE_TABLES.

3) Are there explicit charges for data refresh from offline source table to online table "Snapshot or Triggered sync mode (data transfer)" ?

Re: Online Feature Table : Storage

Walter_C — Tue, 17 Dec 2024 11:40:36 GMT

1) Storage for Online Feature Table's Data:

Online Feature Tables use Databricks' internal storage. They are fully serverless tables that auto-scale throughput capacity with the request load and provide low latency and high throughput access to data of any scale. This means that the data for Online Feature Tables is not stored in the customer's cloud storage but in Databricks' managed infrastructure.

2) Storage Cost Information: For storage cost information, you can refer to the URL provided: Online Tables Pricing. Billing for Online Tables Capacity Units began on or after October 1, 2024, and billing for Online Tables Storage began on or after November 1, 2024. You can use the billing_origin_product = ONLINE_TABLES to get information on capacity units.

3) There is no explicit mention on costs for ofline source table to online operations but let me confirm on this.

Re: Online Feature Table : Storage

AnkurMittal008 — Thu, 19 Dec 2024 05:27:40 GMT

Thanks a lot @Walter_C ..