本頁面將介紹並比較這兩項功能管理服務，並提供其功能總覽。並說明如何將 Vertex AI 特徵儲存庫 (舊版) 中的現有特徵儲存庫遷移至新版 Vertex AI 特徵儲存庫。
Vertex AI 特徵儲存庫
Vertex AI 特徵儲存庫提供全新的特徵管理方式，可讓您從 BigQuery 資料來源維護及提供特徵資料。在這種做法中，Vertex AI 特徵儲存庫會充當中繼資料層，為 BigQuery 中的特徵資料來源提供線上服務功能，並讓您根據該資料線上提供特徵。您不需要將資料複製或匯入 Vertex AI 中的個別離線儲存庫。
Vertex AI 特徵儲存庫已與 Dataplex 通用目錄整合，以追蹤特徵中繼資料。它也支援嵌入，可讓您針對最近鄰執行向量相似度搜尋。
使用 ReadFeatureValues(ReadFeatureValuesRequest) API 進行線上放送。
遷移至 Vertex AI 特徵儲存庫
Vertex AI 特徵儲存庫 (舊版) 的資源和特徵資料無法在 Vertex AI 特徵儲存庫中直接使用。如果您是 Vertex AI 特徵儲存庫 (舊版) 的現有使用者，且想要將專案遷移至 Vertex AI 特徵儲存庫，請執行下列步驟。請注意，由於 Vertex AI 特徵儲存庫中的資源階層與 Vertex AI 特徵儲存庫 (舊版) 中的資源階層不同，因此您必須在遷移特徵資料後手動建立資源。
[[["容易理解","easyToUnderstand","thumb-up"],["確實解決了我的問題","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["難以理解","hardToUnderstand","thumb-down"],["資訊或程式碼範例有誤","incorrectInformationOrSampleCode","thumb-down"],["缺少我需要的資訊/範例","missingTheInformationSamplesINeed","thumb-down"],["翻譯問題","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["上次更新時間：2025-09-04 (世界標準時間)。"],[],[],null,["# Introduction to feature management in Vertex AI\n\nIn machine learning (ML), features are characteristic attributes of an instance\nor entity that you can use to train models or to make online predictions.\nFeatures are generated by transforming raw ML data into measurable and shareable\nattributes using feature engineering techniques, generally referred to as\n*feature transformations*.\n\n*Feature management* refers to the process of creating, maintaining, sharing,\nand serving ML features stored in a centralized location or repository.\nFeature management makes it easier to reuse features to train and retrain models,\nreducing the life cycle of AI and ML deployments.\n\nA product or service that includes feature management\nservices to store, discover, share, and serve ML features is called a\n*feature store*. Vertex AI incorporates the following feature store\nservices:\n\n- [Vertex AI Feature Store](#vaifs)\n\n- [Vertex AI Feature Store (Legacy)](#vaifs_legacy)\n\nThis page introduces and compares the two feature management services and provides\nan overview of their capabilities. It also describes how to migrate an\nexisting feature store in Vertex AI Feature Store (Legacy) to\nthe new Vertex AI Feature Store.\n\n### Vertex AI Feature Store\n\nVertex AI Feature Store offers a new approach to feature management by\nletting you maintain and serve your feature data from a BigQuery\ndata source. In this approach, Vertex AI Feature Store acts as a metadata\nlayer that provides online serving capabilities to your feature data source in\nBigQuery and lets you serve features online based on that data.\nYou don't need to copy or import the data to a separate offline store in\nVertex AI.\n\nVertex AI Feature Store is integrated with Dataplex Universal Catalog to\ntrack feature metadata. It also supports embeddings and lets you perform\nvector similarity searches for nearest neighbors.\n\nVertex AI Feature Store is optimized for ultra-low latency serving\nand lets you do the following:\n\n- Store and maintain your offline feature data in BigQuery,\n taking advantage of the data management capabilities of BigQuery.\n\n- Share and reuse features by adding them to the feature registry.\n\n- Serve features for online predictions at low latencies using Bigtable online serving\n or at ultra-low latencies using Optimized online serving.\n\n- Store embeddings in your feature data and perform vector similarity searches using Optimized online serving.\n\n- Track feature metadata in Dataplex Universal Catalog.\n\nTo learn more about Vertex AI Feature Store, see the\n[Vertex AI Feature Store documentation](/vertex-ai/docs/featurestore/latest/overview).\n\n### Vertex AI Feature Store (Legacy)\n\nVertex AI Feature Store (Legacy) provides a centralized repository to\nstore, organize, and serve ML feature data. It provisions a resource hierarchy\nthat encapsulates both an online store and an offline store within\nVertex AI. The online store serves the most recent feature values\nfor online predictions. The offline store stores and maintains feature data\n(including historical data) that you can batch serve for training ML models.\n\nVertex AI Feature Store (Legacy) is a fully-functional feature management\nservice that lets you do the following:\n\n- Batch or stream import feature data into the offline store from a data\n source, such as a Cloud Storage bucket or a BigQuery source.\n\n- Serve features online for predictions.\n\n- Batch serve or export features for ML model training or analysis.\n\n- Set Identity and Access Management (IAM) policies on `EntityType` and\n `Featurestore` resources.\n\n- Manage feature store resources from the Google Cloud console.\n\nVertex AI Feature Store (Legacy) doesn't include embeddings management\nor vector retrieval capabilities. If you need to manage embeddings in your\nfeature data or perform vector similarity searches, consider switching to\nVertex AI Feature Store. For information about migrating to\nVertex AI Feature Store, see\n[Migrate to Vertex AI Feature Store](#migrate).\n\nTo learn more about Vertex AI Feature Store (Legacy), see the\n[Vertex AI Feature Store (Legacy) documentation](/vertex-ai/docs/featurestore/latest/overview).\n\nComparison between Vertex AI Feature Store and Vertex AI Feature Store (Legacy)\n-------------------------------------------------------------------------------\n\nThe following table compares the various aspects of\nVertex AI Feature Store (Legacy) and the new Vertex AI Feature Store: \n\nMigrate to Vertex AI Feature Store\n----------------------------------\n\nVertex AI Feature Store (Legacy) resources and feature data aren't\nreadily available in Vertex AI Feature Store.\nIf you're an existing user of Vertex AI Feature Store (Legacy) and\nwant to migrate your project to Vertex AI Feature Store, perform\nthe following steps. Note that since the resource hierarchy in\nVertex AI Feature Store is different from the resource hierarchy in\nVertex AI Feature Store (Legacy), you'll need to manually create the\nresources after you migrate the feature data.\n\n1. If your feature data isn't available in BigQuery already,\n [export the feature data to BigQuery](/vertex-ai/docs/featurestore/export-features#export_feature_values),\n and create BigQuery tables and views. Follow the\n [Data preparation guidelines](/vertex-ai/docs/featurestore/latest/prepare-data-source#guidelines)\n when you export and prepare the data. For example:\n\n - Each feature corresponds to a column. Entity IDs can be a separate\n column, which you can identify as the `ID` column.\n\n - Vertex AI Feature Store doesn't have the `EntityType` and `Entity`\n resources. Provide the feature values for each entity in the row\n corresponding to the entity ID.\n\n2. Optional: Register your feature data source by adding feature groups and\n features. For more information, see [Create a feature group](/vertex-ai/docs/featurestore/latest/create-featuregroup)\n and [Create a feature](/vertex-ai/docs/featurestore/latest/create-feature).\n\n3. Set up online serving by creating online store and feature view instances\n based on the feature data.\n\nWhat's next?\n------------\n\n- [Learn more about Vertex AI Feature Store](/vertex-ai/docs/featurestore/latest/overview)\n\n- [Learn more about Vertex AI Feature Store (Legacy)](/vertex-ai/docs/featurestore/overview)"]]
