With Google Cloud Next in full swing, the GC team has launched a couple of new services and previews that look to free teams from infra-maintenance duty — targeting anyone from admins and data scientists through to application developers. If you take a closer look, however, most of the new additions meant to lure users to the platform are powered by one open-source project or another.
Managed machine learning platform Vertex AI, for instance, got a new companion project in the form of Vertex AI Workbench. Still in preview at this point, Workbench is supposed to provide a “single development environment for the entire data science workflow”. While this sounds rather ominous, it boils down to a Jupyter-based product that integrates with Vertex AI for workflow automation via pipelines and can be comparatively easily connected to various data services.
Workbench offers customers a choice between managed notebook instances or a user-managed equivalent of customisable Deep Learning VM image instances, for organisations with specific networking and security requirements. Both flavours come pre-packaged with JupyterLab — a suite of deep learning packages — support for GPU acceleration and synchronisation with GitHub, and are “protected by Google Cloud authentication and authorization”.
In other data-related news, Spark on Google Cloud and multi-cloud data warehouse Big Query Omni have been declared generally available.
PostgreSQL users working with Google Cloud offerings might also be interested to learn that Google started a preview for a PostgreSQL interface for managed relational database Cloud Spanner. The addition is meant to make Spanner an option for users who’ve found it hard to build for in the past, since it only allowed Google SQL as a database dialect. Cloud Spanner is mainly known for its promise of “unlimited scale, consistency, and 99.999% availability” and is therefore especially popular for products in the gaming and finance industries. The new interface is meant to make Spanner more accessible by offering a way to interact with data as if it were stored in a PostgreSQL database.
It realises this by implementing a subset of PostgreSQL data types and DDL syntax — key parts of the PostgreSQL SQL dialect — fitting the clients with capabilities to understand the query dialect, and offering support for the PostgreSQL wire protocol. Before getting too excited though, developers should be aware that the supported syntax is still quite limited and that it might take a while before the interface can offer all the functionality of the original SQL. Google promises that non-functional aspects such as availability and performance should be the same, despite the extra internal query conversion work necessary.
On the operational side of things, Google kicked off a preview for the GC Managed Service for Prometheus this week. Project co-founder Julius Volz saw this as Prometheus coming full-circle, since it was inspired by Google’s internal Borgmon monitoring system in the first place. Google’s service is developed as a drop-in replacement for Prometheus and Thanos, providing fully managed, scalable metrics storage and retrieval on the same data store the company’s Cloud Monitoring product is using.
Other previews of interest are a Log Analytics feature which is meant to help store, manage, and analyse log data, a Cloud Build Hybrid functionality to enable the execution of Cloud Build workloads across cloud and on-premises systems, and Anthos for Virtual Machines. The latter allows admins to connect the control plane of app platform Anthos to vSphere environments, or move VMs onto Anthos via Kubernetes virtualisation API KubeVirt, in order to manage Kubernetes and traditional workloads side by side.
Ops personnel interested in gaining insight into the gross carbon emissions associated with their organisation’s cloud usage can check that via the GCP Cloud Console now. To drive that number down, GC plans to start offering reduction tips via a sustainability impact category in the Active Assist Recommender soon. The first component to provide something along those lines will be the Unattended Project Recommender.