Monitor end-to-end performance beyond the last mile
Fine-grained monitoring tools for applications demanding the highest performance
Identify performance hotspots in the program code.
Use distributed tracing to track critical execution paths in systems built across physical nodes, hardware platforms and coding languages.
Track and monitor key metrics and hardware counters responsible for overall system performance.
Get notified of errors produced by software crashes, hardware events and application exceptions that affect product performance.
Discover bloated code with automatic code profiling
Code-level profiling provides solutions to performance issues that other tools cannot offer. Whether your performance hotspots are in sensors, driver code, GPU code, computer vision algorithms, or machine learning models, profiling can pinpoint the bottleneck down to the function line number. Intuitive interface allows you to drill down modules to get per-function usage breakdowns.
- Profile high-performance C/C++ applications to diagnose memory and CPU usage over its entire lifetime or a specific interval
- Profile GPU code exclusively or alongside CPU execution, including per-stream NVIDIA CUDA kernels
- Profile critical computer vision tasks that run on OpenCV and other CV frameworks
- Profile deep learning models that run on Tensorflow, Pytorch, Keras and other deep learning frameworks
Use distributed tracing to track and measure each step of any execution path and find the bottleneck in your code. Visualize the breakdown of critical tasks and operations by their execution times.
- High-precision nanosecond timestamps for hardware-layer tracing
- Custom tags, metadata, and log messages for quick search and filtering
Real-time performance metrics monitoring
Monitor performance metrics down to hardware counters and registers
Real-time monitoring and logging of performance metrics, including the following:
- Hardware performance counters (eg. CPU, GPU, RAM, SSD, SoC, FPGA, and more)
- System metrics (eg. CPU load, memory usage, disk transfer rate, network throughput)
- Deep learning metrics (eg. accuracy, AUC, logarithmic loss, confusion matrix, RMSE)
- Application framework metrics (eg. Java, .NET, Redis, SQL, Apache, docker, and many more)
Use pre-defined or define custom alerts that trigger when metric values cross a particular threshold.
Error tracking & alerts
Get notifications of errors that are affecting performance
Track errors produced by software crashes, hardware events, application exceptions and alarms that cause service interruptions for your users. Errors can include any of the following:
- Hardware errors (eg. CPU overheat, RAM ECC errors, disk CRC errors)
- Software crashes (eg. segmentation fault, bus error, divide by zero)
Set pre-defined or create your custom alerts when a certain number of errors of a partitcular type is received.