您现在访问的是微软AZURE全球版技术文档网站,若需要访问由世纪互联运营的MICROSOFT AZURE中国区技术文档网站,请访问 https://docs.azure.cn.

什么是 Azure Data Box Edge?What is Azure Data Box Edge?

Azure Data Box Edge 是具有网络数据传输功能的支持 AI 的边缘计算设备。Azure Data Box Edge is an AI-enabled edge computing device with network data transfer capabilities. 本文概述 Data Box Edge 解决方案、其优势、重要功能以及可以部署此设备的场合。This article provides you an overview of the Data Box Edge solution, benefits, key capabilities, and the scenarios where you can deploy this device.

Data Box Edge 是一种硬件即服务解决方案。Data Box Edge is a Hardware-as-a-service solution. Microsoft 为客户提供了一个具有内置现场可编程门阵列 (FPGA) 的云托管设备,该阵列可加速 AI 推理并具有存储网关的所有功能。Microsoft ships you a cloud-managed device with a built-in Field Programmable Gate Array (FPGA) that enables accelerated AI-inferencing and has all the capabilities of a storage gateway.

用例Use cases

下面介绍各种方案,其中 Data Box Edge 可用于加快边缘的机器学习 (ML) 推理并对数据进行预处理,然后再将该数据发送到 Azure。Here are the various scenarios where Data Box Edge can be used for rapid Machine Learning (ML) inferencing at the edge and preprocessing data before sending it to Azure.

  • 使用 Azure 机器学习进行推理 - 借助 Data Box Edge,可以运行 ML 模型以获得可在数据发送到云之前执行的快速结果。Inference with Azure Machine Learning - With Data Box Edge, you can run ML models to get quick results that can be acted on before the data is sent to the cloud. 可以选择传输完整的数据集以继续重新训练并改进 ML 模型。The full data set can optionally be transferred to continue to retrain and improve your ML models. 有关如何在 Data Box Edge 设备上使用 Azure ML 硬件加速模型的详细信息,请参阅在 Data Box Edge 上部署 Azure ML 硬件加速模型For more information on how to use the Azure ML hardware accelerated models on the Data Box Edge device, see Deploy Azure ML hardware accelerated models on Data Box Edge.

  • 预处理数据 - 转换数据,然后将数据发送到 Azure,以创建更具操作的数据集。Preprocess data - Transform data before sending it to Azure to create a more actionable dataset. 使用预处理可以:Preprocessing can be used to:

    • 聚合数据。Aggregate data.
    • 修改数据,例如删除个人数据。Modify data, for example to remove personal data.
    • 用于优化存储和带宽,或用于进一步分析的子集数据。Subset data to optimize storage and bandwidth, or for further analysis.
    • 分析和应对 IoT 事件。Analyze and react to IoT Events.
  • 通过网络将数据传输到 Azure - 使用 Data Box Edge 可以快速轻松地将数据传输到 Azure,以实现其他计算和分析或存档目的。Transfer data over network to Azure - Use Data Box Edge to easily and quickly transfer data to Azure to enable further compute and analytics or for archival purposes.

关键功能Key capabilities

Data Box Edge 具有以下功能:Data Box Edge has the following capabilities:

功能Capability 说明Description
加速 AI 推断Accelerated AI inferencing 通过内置 FPGA 实现。Enabled by the built-in FPGA.
计算Computing 允许分析、处理、筛选数据。Allows analysis, processing, filtering of data.
高性能High performance 高性能计算和数据传输。High performance compute and data transfers.
数据访问Data access 使用云 API 从 Azure 存储 Blob 和 Azure 文件中直接访问数据,以便在云中进行其他数据处理。Direct data access from Azure Storage Blobs and Azure Files using cloud APIs for additional data processing in the cloud. 设备带有本地缓存,以便快速访问最近使用的文件。Local cache on the device is used for fast access of most recently used files.
云托管Cloud-managed 设备和服务通过 Azure 门户进行管理。Device and service are managed via the Azure portal.
离线上传Offline upload 离线模式支持离线上传方案。Disconnected mode supports offline upload scenarios.
支持的协议Supported protocols 支持用于数据引入的标准 SMB 和 NFS 协议。Support for standard SMB and NFS protocols for data ingestion.
有关支持的版本的详细信息,请转到 Data Box Edge 系统要求For more information on supported versions, go to Data Box Edge system requirements.
数据刷新Data refresh 可以使用云中的最新内容刷新本地文件。Ability to refresh local files with the latest from cloud.
加密Encryption BitLocker 支持本地加密数据,并通过 http 安全地将数据传输到云中。BitLocker support to locally encrypt data and secure data transfer to cloud over https.
宽带限制Bandwidth throttling 中止以限制在高峰时段使用带宽。Throttle to limit bandwidth usage during peak hours.

组件Components

Data Box Edge 解决方案包括 Data Box Edge 资源、Data Box Edge 物理设备和本地 Web UI。The Data Box Edge solution comprises of Data Box Edge resource, Data Box Edge physical device, and a local web UI.

  • Data Box Edge 物理设备 - 可将 Microsoft 提供的 1U 机架安装式服务器配置为向 Azure 发送数据。Data Box Edge physical device - A 1U rack-mounted server supplied by Microsoft that can be configured to send data to Azure.

  • Data Box Edge 资源 - Azure 门户中的一个资源,使用该资源可以通过 Web 界面(可从不同的地理位置访问该界面)管理 Data Box Edge 设备。Data Box Edge resource – a resource in the Azure portal that lets you manage a Data Box Edge device from a web interface that you can access from different geographical locations. 使用 Data Box Edge 资源可以创建和管理资源、查看和管理设备与警报,以及管理共享。Use the Data Box Edge resource to create and manage resources, view, and manage devices and alerts, and manage shares.

    有关详细信息,请转到为 Data Box Edge 设备创建订单For more information, go to Create an order for your Data Box Edge device.

  • Data Box 本地 Web UI - 使用本地 Web UI 可以运行诊断、关闭和重启 Data Box Edge 设备、查看复制日志,并联系 Microsoft 支持部门来提出服务请求。Data Box local web UI - Use the local web UI to run diagnostics, shut down and restart the Data Box Edge device, view copy logs, and contact Microsoft Support to file a service request.

    有关使用基于 Web 的 UI 的详细信息,请转到使用基于 Web 的 UI 管理 Data BoxFor information about using the web-based UI, go to Use the web-based UI to administer your Data Box.

上市区域Region availability

将数据传输到的 Data Box Edge 物理设备、Azure 资源和目标存储帐户不一定非要位于同一区域。Data Box Edge physical device, Azure resource, and target storage account to which you transfer data do not all have to be in the same region.

  • 资源可用性 - 有关 Data Box Edge 资源可用的所有区域的列表,请转到可用的 Azure 产品(按区域)Resource availability - For a list of all the regions where the Data Box Edge resource is available, go to Azure products available by region. Data Box Edge 也可以部署在 Azure 政府云中。Data Box Edge can also be deployed in the Azure Government Cloud. 有关详细信息,请参阅什么是 Azure 政府?For more information, see What is Azure Government?.

  • 目标存储帐户 - 存储数据的存储帐户可在所有 Azure 区域中获得。Destination Storage accounts - The storage accounts that store the data are available in all Azure regions. 存储帐户存储 Data Box Edge 数据的区域应靠近设备所在位置,以便获得最佳性能。The regions where the storage accounts store Data Box Edge data should be located close to where the device is located for optimum performance. 远离设备的存储帐户会导致长时间的延迟和性能下降。A storage account located far from the device results in long latencies and slower performance.

后续步骤Next steps