Snowflake is introducing several FinOps innovations:
- Cost-Based Anomaly Detection: (Public preview) Now integrated into the Cost Management interface (which is now generally available for roles other than ACCOUNTADMIN), it alerts users to abnormal cost spikes.
- Tag-Based Budgets: (Generally available soon) Allows establishing spending limits based on resource tags.
- Enhanced Performance Visibility: Tools like Performance Explorer (GA soon), Query Insights (private preview), and Cost Insights for Large Query Gaps (private preview) offer better optimization.
- FinOps Foundation Membership: Snowflake has joined as a Premier Enterprise Member to help drive and adopt industry best practices.
Snowflake Adaptive Compute (in private preview) is the next evolution of Snowflake's compute service. It automates infrastructure choices by automatically selecting the appropriate cluster size(s), number of clusters, and auto-suspend/resume duration for jobs. It also intelligently routes queries to right-sized clusters without user action, saving time, minimizing configurations, and improving price/performance. Warehouses created using this service are called Adaptive Warehouses.
Standard Warehouse – Gen2 is an updated version of Snowflake’s current Standard Warehouse. It features upgraded hardware and additional performance enhancements, delivering 2.1x faster performance for core analytics workloads. This is generally available.
The announcements focus on making the Snowflake platform even easier to use, more performant, and more trusted. Key themes include intelligent compute innovations, enhanced FinOps capabilities, easier cross-account management, an AI-powered catalog for governance and discovery, comprehensive security enhancements, improved observability, robust business continuity/disaster recovery, greater extensibility, and new ways to leverage transactional data.
Many organizations struggle with the complexity and cost of managing diverse data, applications, and models. They often use piecemeal solutions from multiple cloud vendors, leading to manual configurations, difficult governance, security challenges, wasteful downtime, and underutilized data. Snowflake provides a fully managed platform that automates administration, scales with needs, integrates AI with data, and offers continuous cost savings through automatic performance improvements and built-in FinOps.
Canvas is available in beta to ChatGPT Plus and Team users globally. Enterprise and Edu users will get access a week later, and it's planned to be available to all ChatGPT Free users once it exits beta.
Canvas is a new interface in ChatGPT designed for working on writing and coding projects. It allows users to collaborate with ChatGPT on projects beyond simple chat, providing tools for editing, refining, and managing ideas within a separate window.
Snowflake’s ML functions are subject to review and oversight, with outputs possibly containing inaccuracies or biases.
A gradient boosting machine algorithm that models trends, auto-regressive lags, and cyclic calendar variables, handling features without one-hot encoding.
Factors include the size and number of rows and columns in your data, with recommendations to use Snowpark-optimized warehouses for larger datasets.
The !EXPLAIN_FEATURE_IMPORTANCE method ranks features based on their impact on predictions, showing their importance on a normalized scale.
Use the model!SHOW_EVALUATION_METRICS method to view metrics like MAE, MAPE, and SMAPE for evaluating model performance.
Include columns like weather data in the model; these features influence predictions based on trends in those data points.
Combine identifiers like store_id and item into a new column and use it with the SERIES_COLNAME parameter for multi-series forecasts.
Yes, Snowflake's model training is designed to manage real-world data issues, including missing, duplicate, or misaligned time steps.
Data should include columns like store_id, item identifier, timestamp, and sales values, potentially with additional features like temperature and holiday indicators.
Forecasting can handle multiple series, use additional features, assess model accuracy, and understand feature importance.
Use CREATE SNOWFLAKE.ML.FORECAST with appropriate parameters to train a forecasting model and start generating predictions.
A database, schema, and virtual warehouse; CREATE SNOWFLAKE.ML.FORECAST privileges; a table or view with at least one timestamp column and one numeric column with fixed intervals.
Forecasting uses machine learning to predict future numeric data based on historical data, commonly for cases like predicting sales over a specified period.
