Developing and operationalizing complex business logic and ML algorithms at cloud-scale
Xcalar Data Platform provides the ability to work relationally with petabytes of data, while applying ML algorithms into the mix of techniques used. Users model and operationalize their business logic using a combination of relational, visual, and structured programming paradigms along with machine learning for industry’s lowest time-to-value.
Today’s modern day distributed applications and business logic need relational programming with the ability to apply join, union, group by, pivot, filter, aggregate, sort and merge operations, structured programming with the ability to do loops and conditionals, statistical programming, numerical analysis, and most importantly ML/AI algorithms. While several tools and libraries make it possible to build ML/AI models on a laptop or a cloud VM, operationalizing these models on real-world data at the scale of billions of classifications in seconds is hard. Businesses find that they must provision enormous compute resources, armies of specialized engineers, and inordinate amounts of time in order to process petabytes of data.
With Xcalar, organizations can expect to reduce data processing operating costs to 1/10th of their existing costs - a huge 90% reduction!
Data is ephemeral, but the information within is not. Businesses have a limited time window to get to it. Establishing causal relationships and extracting useful information from data in a cost efficient and timely fashion requires using a combination of techniques - including complex relational compute and ML. Xcalar provides the ability to use a variety of different methods to get actionable business insights with high accuracy.
The three approaches often used by users to predict answers, involve:
- Pure structured programming (touching all the data points to find precise answers, which can become impractical as the data grows to petabyte scale)
- Statistical ways (sampling, to get quicker answers)
- ML (to find patterns, make inferences)
Xcalar gives users the ability to use a combination of all of these approaches.
With Xcalar, users build models with any open source tool of their choice and then operationalize them at cloud-scale with enterprise-grade reliability and high performance. Xcalar provides the simplicity, speed, and scale for users to get results rapidly, and with significant cost savings.
Ad hoc analytics/modeling
Responsive interface performs interactive analysis using relational operators on up to 100 billion rows.
Powerful SQL paradigm
Users can work with standard ANSI SQL applying thousands of operations like join, union, group by, pivot, filter, aggregate, sort and merge operations, in series or in parallel.
Visual programming with lineage
Users work interactively with very large diverse datasets as virtual tables to create dataflow models. These models track lineage of data from source through each transformation.
Operational machine learning
Data scientists can train and deploy ML algorithms across petabytes of data at any stage of the data pipeline.
Users apply proprietary logic, including loops and conditionals, to dataflow models, using Python.
Exceptional scalability and performance
Xcalar Data Platform processes both read and write operations across cloud-scale clusters with near-linear scalability while maintaining strong data consistency.
Handling Data Anomalies
While processing each data operation, Xcalar Data Platform surfaces the anomalies inherent in the data; users can triage these at any stage of analytics work.