The GDC AI search solution empowers organizations to deploy an on-prem conversational search tool that uses natural language processing to find the most relevant information from their data. This can significantly boost employee productivity and knowledge sharing, as search queries and data remain on-prem, ensuring compliance with data sovereignty regulations. The architecture of this solution, as illustrated in the attached diagram, integrates various AI and data processing tools:
- Data Ingestion: OCR models, translation models, and speech-to-text models from Vertex AI convert images, audio, and text into English text.
- Embedding and Indexing: The text is then embedded into vectors using the E5 Large V2 model and indexed in a vector database (AlloyDB Omni).
- Query Processing: When a user submits a query via a chat interface, it is embedded into a vector and searched against the indexed vectors in AlloyDB Omni.
- Result Retrieval and Response Generation: The relevant document chunks are retrieved and sent to a large language model (LLM) like Gemma 7B, which generates an AI-driven response.
This solution provides a comprehensive AI stack on-premises, integrating Google-developed models, third-party solutions, and open-source tools. It supports various pre-trained APIs for tasks such as speech-to-text, translation, and optical character recognition, ensuring versatile data handling capabilities.
Security and Platform Enhancements
Deploying AI and other enterprise-grade applications demands a robust, secure platform. GDC places a strong emphasis on security and compliance, enabling organizations to meet stringent regulatory requirements while protecting sensitive workloads. Several new security and platform enhancements have been introduced:
Security Enhancements
- ISO27001 and SOC2 Compliance: GDC has achieved these certifications, demonstrating its commitment to meeting high standards of regulatory and compliance requirements.
- Managed Intrusion Detection and Prevention Solution (IDPS): This integrates Palo Alto Networks' threat prevention technology with the GDC architecture, inspecting north-south traffic and providing transparent inline protection for GDC workloads.
Platform Enhancements
- GDC Sandbox:This managed environment allows customers and partners to build and test services for GDC in a Google Cloud environment, bypassing the need for physical hardware.
- GDC Racks:The latest-generation racks are optimized for AI and general compute workloads, offering flexibility with network- or storage-optimized nodes.
- Storage Flexibility:GDC now supports independent growth of storage to accommodate large analytics or AI workloads, with options for block, file, or object storage.
- Survivability Enhancements:GDC supports disconnected mode for up to seven days and includes offline management features to ensure continuous operation even when disconnected.
- Apigee Support:Google Cloud's API management solution, Apigee, is now supported on GDC, enabling scalable API management for AI workloads.
GDC Ecosystem with Nvidia
Collaboration with partners is key to deploying AI applications effectively at the edge or on-prem. Google continues to expand the GDC ecosystem, working closely with partners to provide specialized skills and technologies. Notable developments include:
- Managed GDC Providers: Partners such as Clarence, T-Systems, and WWT now offer GDC as a managed service, providing end-to-end solutions including operation, design, migration, and deployment services.
- Google Cloud Ready — Distributed Cloud Badge: This new certification helps customers identify independent software vendors (ISVs) whose solutions are optimized for GDC. Partners like Canonical, CIQ, Elastic, MongoDB, Palo Alto Networks, and Starbust have already validated their solutions, with more partners in the pipeline.
- Nvidia Partnership: Google continues to collaborate with Nvidia to integrate the latest GPUs for AI workloads. The new GDC server features the energy-efficient Nvidia L4 Tensor Core GPU, designed for use cases where high availability is not required, alongside the AI-optimized servers with Nvidia H100 Tensor Core GPUs.
The enhancements to Google Distributed Cloud are poised to revolutionize how businesses deploy and manage AI workloads across diverse environments. By bringing Google's AI capabilities closer to where data is generated, GDC ensures compliance, enhances security, and provides the flexibility needed to innovate at scale. With new generative AI solutions, robust security features, platform enhancements, and a growing ecosystem of partners, GDC empowers organizations to fully leverage the power of AI, wherever they need it.
Reference: Google Cloud Blog | Run AI anywhere with Google Distributed Cloud innovations
