AI Management Console
[:tw][vc_row equal_height=“yes” parallax=“content-moving” parallax_image=“2559” parallax_speed=“4” parallax_x=“4” parallax_y=“4”][vc_column][vc_single_image image=“1223” img_size=“medium” alignment=“center”][title-section style=“style2” title=“AI Management Console” class=“fcolorfff”][/title-section][vc_empty_space][/vc_column][/vc_row][vc_row][vc_column][vc_column_text]
產品簡介
從大數據的資料預處理、AI 模型的訓練、一路最終至 AI 預測應用部署的複雜流程中，企業經常因不同階段所需的異質環境、多種運算架構、甚至跨部門的協作而困擾。透過 Gemini AI console 打造出便利企業組織進行跨單位 AI 專案協作的 GPU 管理節點，協助企業更有效率地從海量資料挖掘出更好的商機。 資料科學家與開發人員可透過 Gemini AI console 使用者服務入口，便利地快速開啟大量預載大數據與 AI 工具的運算叢集環境，將底層運算架構建置化繁為簡，只需透過網頁表單點選即可立即開啟各種所需的AI運算工具服務，進而能夠將時間與人力資源專注投入在核心演算法上。[/vc_column_text][/vc_column][/vc_row][vc_row][vc_column][vc_single_image image=“2362” img_size=“full” alignment=“center”][/vc_column][/vc_row][vc_row][vc_column][vc_column_text]
| 功能 | 描述 |
| 權限與角色管理 | 系統管理者可建立多個Project(群組)，並管理使用者帳號與權限，系統管理者可利用LDAP或內建認證機制來確認哪些使用者可以使用哪些服務，並提供工作流程給使用者申請及管理者審核管理。 |
| AI服務架構制定管理 | 系統管理者可根據GOC格式語法描述雲端服務並上架到GOC PaaS，支援兩種運算架構制定：雲服務叢集 (Cloud Service Clusters) 與批次工作(Batch Jobs)；使用者及管理者可透過Image Registry自行上傳及使用特定的映像檔。 |
| 資源與排程管理 | GOC為AI服務提供自助化虛擬資源的佈署，並根據不同AI服務的需求可運行於異質環境(VM, Docker, GPU)，在批次工作模式下，依使用者預約或派送的順序進行工作排程管理，此外，GOC支援群組資源配額管理及水平橫向擴充(Scale-out)。 |
| 整合企業既有服務架構 | GOC 架構可讓企業輕鬆整合既有的系統服務 如 LDAP, NFS 等。因此 現有 使用者可 log in 到 GOC，去直接使用 存在 NFS server 上的檔案。此外，企業 也可以 沿用既有 的 組織結構權限列表 (ACL) 去控管 使用者 的 權限 跟 數據。 |
| 資源統計與監控 | 透過使用者儀表板介面呈現所有雲端服務(Cloud Service)與批次工作(Batch Job)的資源即時使用狀況及歷史資料，並可提供匯出使用資源報表作為計費基礎。 |
| 服務管理介面 | 提供兩個入口平台給系統管理者與使用者進行資源管理，包括讓使用者自行輸入簡易參數申請建立新的雲服務(create site)與派送工作(submit job)，使用者可透過儀表板監看服務與工作資訊，並提供管理者異常狀況自動告警通知。 |
產品效益
1.簡化IT複雜性，優化GPU管理效益：Gemini AI Management Console 幫助管理者在單一平台上管理從單台至數百台的GPU 及CPU Server的實體與虛擬資源
2.提升研究人員研發效率，縮短開發時間：
將底層運算架構建置化繁為簡，只需透過網頁表單點選即可立即開啟各種所需的巨資與AI運算工具服務，使用者得以專注在分析演算法開發
3.支援不同運算架構與異質環境，完美AI運算體驗：
單一平台同時管理大數據+AI機器學習運算叢集，無論底層是透過虛擬機(VM)或是容器(Docker)來搭配GPU，都能讓使用者一目瞭然、輕鬆管理，並支援雲端服務及批次工作模式，滿足不同的使用需求情境，提升AI開發流程體驗
[/vc_column_text][vc_empty_space][/vc_column][/vc_row][vc_row][vc_column][vc_single_image image=“2612” img_size=“full” alignment=“center”][/vc_column][/vc_row][vc_row][vc_column][vc_row_inner][vc_column_inner][vc_column_text]
- 單一入口Web介面提供DevOps使用者雲服務與AI工具佈署管理操作
- 單一入口Web介面提供IT管理者針對資源與使用者進行管理，並提供雲市集架構讓 IT能快速將雲服務上、下架，支援企業快速開發，上架 AI 應用流程
- 多雲資源管理：支援 VM、容器、物件儲存及檔案儲存，並可分成專屬區與分享區兩大類型資源管理
- 提供完整Restful API，搭配API Gateway可提供客製化需求
Introduction
The development of artificial intelligence (AI) undoubtedly brings unprecedented technological progress in human history. Gemini AI Console helps data engineers to manage single or multiple GPU servers easily, allowing valuable GPU resources to be used more efficiently, further reduces the cost of Big Data and AI development. From big data preprocessing, AI model training, all the way to the complex process of AI application deployment, enterprises often suffer from heterogeneous environments, multiple computing architectures, and even cross-departmental collaborations required at different stages. Gemini AI console aims to facilitate enterprise and organization for cross-unit AI project collaboration, therefore help enterprises more efficiently extract better business opportunities from massive data.[/vc_column_text][/vc_column][/vc_row][vc_row][vc_column][vc_single_image image="2362" img_size="full" alignment="center"][/vc_column][/vc_row][vc_row][vc_column][vc_column_text]Features
| Features | Description |
| User and Roles Management | System administrator can create multiple projects and manage user accounts, roles and authorization. System administrator can use the Corporate LDAP or other authentication mechanisms to authorize users for specific services. |
| Jobs/Service Definition | Gemini AI Console supports two kinds of operation models: Cloud Service Clusters (pay-by-capacity) and Batch Jobs (pay-by-utility). Users and administrators can upload and use specific VM and container image files through our private Image Registry. |
| Resource Management and Workload Management | Gemini AI Console provides user self-provisioning for AI services which can run in heterogeneous environments (VM, Docker and GPU). In batch mode, jobs will be scheduled by the GOC workload manager. In addition, Gemini AI Console supports resource quota management by project and horizontal scale-out. |
| Integrate with Existing Service Architecture | GOC architecture allows organizations to easily integrate existing system such as LDAP, NFS, and more. Therefore, existing users can log in to Gemini AI Console to directly use the file on the NFS server. In addition, companies can use existing organizational structure access lists (ACLs) to control user permissions and data. |
| Resource Monitoring and Reporting | Through the user dashboard interface, the real-time usage statistics and historical data of all cloud services and jobs are presented, and the resource usage report can be provided for the billing basis. |
| Management Portal | Gemini AI Console provides a user friendly portal for both system administrator and users to manage resources and services, including creating new service or submit jobs with few clicks. This management portal also provides automatic alarm notification for abnormal events. |
Benefits
- Simplify IT complexity and optimize GPU management Manage the physical and virtual resources for multiple GPU and CPU Servers with a single platform. It can also optimize the utilization of GPU resources according to the needs of business organization and cross-unit AI projects.
- Improve R&D efficiency and shorten development time Make it easier to prepare the complex infrastructure environment with simple browser interface for the deployment of Big Data and AI computing tools, which helps scientists to focus on their AI algorithm development and training.
- Support different computing architectures and heterogeneous environments, brining perfect AI computing experience A single platform can manage big data services and AI machine learning clusters at the same time. No matter using virtual machines (VMs) or containers (Docker) with GPU, users can manage easily through the management portal. Support both ‘Cloud Service’ and ‘Batch Jobs‘, which can meet different usage scenarios and definitely enhance the AI development process experience.
- Single Web portal for both IT operation team and AI development team
- The single-entry web interface provides IT managers to manage resources and users, and to provide a cloud marketplace architecture that allows IT to quickly and dynamically deploy AI cloud services
- Multi-cloud Management: support VM, container, object storage and file storage, and can be divided into two types of computing resources: dedicated zone and shard zone.
- Provide complete Restful APIs with API Gateway for customized requirements