Built-in Trending and Predictive Behavioral Analytics
MDTSDB - it's a mouthful, but these 6 letters will change the way you look at data...
Fractal Industries' Multidimensional Time Series Database (MDTSDB) is a cutting-edge, distributed database system with the flexibility to support data ingestion from virtually any data model. An intuitive interface allows users to configure new data sources or existing datasets to integrate seamlessly with the rest of their data—in minutes instead of weeks or months (if ever). And because MDTSDB automatically indexes time series data with geospatial information, new and existing data is fundamentally enriched with vital context to reveal hidden correlations and deeper insights.
By storing data as time series, MDTSDB enables real-time anomaly or event detection, as well as behavioral forecasting based on advanced pattern matching and trend analysis. Embedded support for event-condition-action (ECA) rules allow triggered behaviors such as notifications or other automated processes as soon as user-defined conditions are met.
A powerful query language delivers built-in capabilities ranging from ad hoc queries, to advanced “fuzzy” search, to complex statistical analytics including feature extraction, clustering, modeling, interpolation, and regression forecasting. And its distributed and parallel architecture also enables querying, processing, and analytics that scale horizontally and can improve response times by a factor of 10 or 15 over traditional scripted queries.
Example Use Cases
Rapid Data Ingest
MDTSDB provides a single interface to configure, consume, correlate, and manage data feeds from almost any internal or external source, allowing users to configure the ingestion of a new data feed in minutes, instead of days or weeks of manual configuration and integration.
Seamless Data Integration
In addition to easily ingesting new data feeds, MDTSDB’s intuitive query language automatically enables integration and correlation of widely disparate data sources and data types, using out-of-the-box or custom parsers to feed a unified, contextualized, and easily queried “data lake.”
Event-Condition-Action (ECA) Automation
Monitor any data stream and trigger actions such as notifications or automated processes when anomalies or other user-defined conditions are met to drive optimized automation for workflows across domains from cybersecurity and quantitative trading to manufacturing, marketing, and operations.
Leverage advanced, multivariate statistical modeling techniques to discover and track unanticipated correlations and interactions between complex system components and processes over time, continuously driving Business Process Management (BPM) and Key Performance Indicator (KPI) metrics toward optimization.
Turn-Key Integration with Fractal OS Components
The full scope of MDTSDB’s capability is realized when deployed with the end-to-end analytic elements of Fractal’s other components— combining foundational data handling, analytics, and automation with next-generation simulation modeling and deep learning to achieve a feedback loop of continuous improvement for virtually any conceivable workflow.
Unmatched Query Capabilities
Ad Hoc Queries
Build customized, ad hoc queries of datasets, choose from a variety of rich visualization options to correlate, sort, and render results in seconds, and store these queries for future reference or automation.
Advanced Fuzzy Search
Baked-in, powerful matching capabilities including phrases, wildcards, joins, grouping and more seamlessly enable discovery of terms that are similar to a specified term without being an exact match.
Geospatial data (e.g. latitude and longitude) is automatically indexed if available, enabling users to compare and evaluate potential spatial relationships, symmetries, and trends over time using queries involving geometric shapes and detailed mapping data.
As a distributed time series database, MDTSDB is horizontally scalable and therefore can perform high-speed, parallel write operations on a cluster, and can answer multidimensional analytical queries in parallel.
Advanced Statistical Analytics
Embedded query language capabilities support Kernel Density Estimation as well as time series analytics including feature extraction, clustering, modeling, interpolation, and regression forecasting.
As more explicit data is collected over time, machine learning algorithms perform pattern matching and trend analysis to establish relevant and increasingly precise benchmarks to immediately detect, triage, and alert on anomalous behavior or significant changes involving any entity or process in the enterprise.
Predict and optimize the performance of a user, device, or application based on recursive pattern matching and trend analysis of its behavior under varying workloads.
Quickly and easily quantify the effect that changes made to an application have on conversion rates down the line.
Lifetime Value of Customers
Discern lifetime ROI of customers through customizable queries to isolate trends and track effectiveness of behavioral queues.
Data Source Health Monitoring
Detect anomalies, protect against silent failures, and drive optimization by continuously monitoring and analyzing the behavior, health, and volume of data feeds over time.
• Code is written entirely in Erlang – state-of-the-art, concurrent, distributed, and fault tolerant
• Incoming data streams per individual host/client system process are lightweight, isolated processes—able to write new data in soft real-time without blocking on writes, with fine-grained access control
• Multidimensional swimlanes provide unprecedented flexibility that supports virtually any data model
• New data sources configurable for integration in minutes instead of weeks or months
• Supports HTTP, protobuffers, AMQP, and websockets, which can be lower overhead and higher speed
• Nanosecond-level time series data along with automatically indexed geospatial data and user-defined tags support advanced but easily defined and context-rich geospatial-temporal queries and analysis across user-created categories
• Specialized API endpoints and indexing for dense sensor data and sparse (or periodically inconstant) event data to maximize query performance
• Data source health monitoring detects anomalies and protects against silent failures
• SQL-like query language that also supports imperative and functional-style programming languages, easily enabling complex queries which leverage mathematical models and methods for advanced statistical analysis
• Powerful matching capabilities, including phrases, wildcards, joins, grouping and more, seamlessly enable discovery of terms that are similar to a specified term without being an exact match
• Indexes on whole value of the data point, on a field of the data point, and on some aggregation functions of the data point value (like EXIST, AVG, COUNT, etc.) to enable fast, efficient, and horizontally scalable queries and analytics
• Avoids sending multiple commands to fetch the desired data due to the rich query language, enabling server-side data processing during ingest and at query time
• Event-Condition-Action (ECA) rules are embedded into MDTSDB query language - a user may ask triggers to fire on data ingest, or create scripts to analyze data after they are stored
• Can be integrated with Fractal's DCG Pipeline Builder and Rules System for even more complex queries and workflows
• External notifications and alerting support achieved via Fractal's Connector component
• Embedded query language capabilities support Kernel Density Estimation as well as time series analytics including feature extraction, clustering, modeling, interpolation, and regression forecasting
• Ad hoc query capabilities allow users to choose from a variety of rich visualization options to correlate, sort, and render results in seconds, and store these queries for future reference or automation
• Supports integration of virtually any external tooling via APIs
• Flexible swimlane-level data expiration, as well as support for automatic transfer of expired data to S3 or Glacier with the option to asynchronously restore it when needed