Alibaba Cloud ARMS RUM vs Zymtrace

Alibaba Cloud ARMS RUM has no write-up on Specifiction yet; its figures are above.

Zymtrace profiles CPU and GPU workloads to help teams find performance bottlenecks in AI training and inference. Its whole-system CPU profiler supports languages including Python, Java, Go, Rust, C/C++, Node.js, Ruby,.NET, PHP, Perl, Zig, and Scala. GPU tools correlate CPU and GPU execution, inspect CUDA kernels, GPU instructions, and stall reasons, and monitor measures such as utilization, memory, power, temperature, SM efficiency and occupancy, Tensor Core utilization, and data transfer rates. The GPU profiling page lists vLLM, Ollama, llama.cpp, and SGLang as supported inference engines. Zymtrace runs on Linux; the maker says eBPF dependence currently rules out Windows. All plans are self-hosted, and the maker says customer data stays within customer infrastructure. The air-gapped option performs processing and analysis locally. Instances provide a REST API, and profiling data can be exposed through MCP for AI agents. OpenTelemetry resource attributes and standard-format export are supported. Basic is free, billed Free Forever, and covers CPU profiling for up to five concurrent machines; GPU profiling is available as a trial.