A new feature called delayed destruction has been introduced, which performs a soft-delete on a secret version when a destroy action is taken. The version is moved to a disabled state and remains in an archival period for a configurable number of days, during which an administrator can revive it. After the delay period expires, the secret version is permanently destroyed. A Pub/Sub notification has been added to inform administrators and SecOps specialists about attempts to destroy a secret version.

Source: https://cloud.google.com/blog/

Google Cloud Functions has been rebranded as Cloud Run functions, offering developers access to all of Cloud Run's capabilities, including GPU support for Python developers using Hugging Face models. Cloud Run's scale-to-zero and fast startup capabilities are ideal for developers starting with AI using Hugging Face models. Existing Cloud Functions 2nd gen functions will be automatically converted to Cloud Run functions, with continued support for existing APIs, gcloud commands, and Terraform modules.

Source: https://cloud.google.com/blog/

Cloud Run now supports NVIDIA L4 GPUs in preview, enabling developers to host real-time inference apps with open gen AI models. This allows for use cases such as performing real-time inference with lightweight open models, serving custom fine-tuned gen AI models, and scaling to handle spiky user traffic. Google's open Gemma (2B/7B) models and Meta's Llama 3 (8B) are examples of models that can be used with Cloud Run.

Source: https://cloud.google.com/blog/

BigQuery introduces TreeAH index, an optimized vector index for large queries, building on its AI-ready data platform capabilities. This new index type uses Google's research and innovation in approximate nearest neighbor algorithms. The TreeAH index is part of BigQuery's efforts to empower customers to build pipelines and applications, such as semantic search and LLM-based retrieval-augmented generation.

Source: https://cloud.google.com/blog/

Bigtable can streamline analytics workflows by tracking billions of interactions with content daily, such as impressions, views, and clicks, without the need for expensive queries and delayed metrics.

Source: https://cloud.google.com/blog/

Web-based large-language models (LLMs) are revolutionizing online interactions, but their integration can introduce new vulnerabilities such as prompt injections and insecure output handling. Security teams are recommended to update their approach to assessing and adapting existing security methodologies for LLMs.

Source: https://cloud.google.com/blog/

Google Security Operations is evolving its converged portfolio, with customers like Etsy making significant changes in their security operations. Manan Doshi, senior security engineer at Etsy, believes this is an opportunity to reset and start anew. The SIEM market is undergoing shifts, causing security operations leaders to rethink their platforms and approach.

Source: https://cloud.google.com/blog/

Apache Airflow is a popular tool for orchestrating data workflows, and Google Cloud offers a managed Airflow service called Cloud Composer. Cloud Composer is a fully managed workflow orchestration service built on Apache Airflow that enables users to author, schedule, and monitor pipelines. A robust logging and alerting setup is crucial to monitor DAGs and minimize downtime in data pipelines.

Source: https://cloud.google.com/blog/

Google announces the general availability of the C4 machine series, a general-purpose VM for Compute Engine and Google Kubernetes Engine customers. The C4 VMs provide industry-leading performance, with up to 20% better price-performance for general-purpose workloads and 45% better price-performance for CPU-based inference compared to other leading hyperscalers. The C4 series is engineered from the ground up and fine-tuned to deliver enterprise-grade reliability and granular control.

Source: https://cloud.google.com/blog/

Google Cloud announced Spanner Graph in preview, a graph database capability for Spanner, and Spanner full-text search, which allows searching for words, phrases, or integers in tables. The announcements were made at Google Cloud Next in Tokyo. Spanner Graph and full-text search capabilities are now available for preview.

Source: https://cloud.google.com/blog/