Theseus Query Profiler
Always on. Negligible overhead. GPU-aware.
Theseus Monitor
Lightweight, fast, always on.
Theseus has an advanced distributed scalable performance monitoring and observability system based on Finite State Machine (FSM) models of engine components. It captures states, transitions and many metrics of operators and tasks, and it tracks all data movement across the engine. The models provide deep insight into the performance of the engine running GPU-accelerated queries, which can help you optimize your queries.
Query Visualization
Theseus' Query Profiler app visualizes your query, allowing you to explore the query plan, how the engine executed it, and when data moved between workers and GPUs.
Customize how the plan is rendered. Overlaying any metric on operator nodes and data flow edges. Need things to go at hyperspeed? Deep-dive into the task execution and data movement timelines and look under the hood for things to tune.
Directed Acyclic Graph (DAG) View
Visualize the query plan. Compare planner predictions and measurements.
Detailed Inspection
Click, or hover over, any node to see its detailed metrics and performance data.
Data In, Data Out
Visualize the amount of data being processed and transferred at every step of the query.
Memory & Interface Bandwidth Consumption
Theseus' Query Profiler app visualizes memory and interface bandwidth utilization across the entire execution of your query. Monitor resource consumption and identify potential bottlenecks before they impact performance.
Memory Utilization
Track memory utilization across multiple devices (e.g. GPU device memory, Host CPU main memory, Storage).
Interface Bandwidth
Monitor data movement between different device memories (GPU to GPU, GPU to Host, Host to Storage, etc), and between workers in the cluster.
Understand bottlenecks
Identify peak usage of your infrastructure and understand how bottlenecks form.
Task Execution Drill Down
Theseus' Query Profiler app provides comprehensive task-level insights, allowing you to analyze even the smallest chunks of computation that were performed for your queries, without overhead. Deep-dive on task states, resource utilization, and performance characteristics.
Task Execution Timeline
Track the asynchronous task execution mechanism that interacts with the GPU.
Task Breakdown
Understand task execution stages and see them run on executor threads.
Task Metrics
Inspect the amount of memory reserved, the amount of data and number of rows processed, and the amount of time spent for every stage of a task.
Ready to Experience GPU-Accelerated SQL?
Start using Theseus today and see the difference GPU acceleration can make in your data processing workflows.