Create benchmarking.md

[DataBoyTX]

detail the benchmarking process to accurately size a system for various use cases

[lmeyerov]

Thanks!

Let's land this, and separately, add forward/backwards in the other capacity/deploy planning links to point to this one as well?

[lmeyerov]

cc @aucahuasi if anything you recommend to add, e.g., any pointers to any perf measurement usage guides

[aucahuasi]

cc @aucahuasi if anything you recommend to add, e.g., any pointers to any perf measurement usage guides

We can mention that telemetry traces are useful for Step 5: Analyze Results:

Use Telemetry Traces: Utilize telemetry traces to gain deeper insights into system behavior and performance bottlenecks. Graphistry supports OpenTelemetry, enabling comprehensive performance monitoring and analysis.`

Also, we can hint the user about testing if they have GPUDirect Storage support, something like:

Additional Step: Testing GPUDirect Storage Support
Check if your system supports NVIDIA GPUDirect Storage by verifying GPU and RDMA-capable adapter compatibility, ensuring drivers and software are installed, and validating GDS with benchmarking tools.  Graphistry supports RAPIDS cuDF for ETL processes.

[DataBoyTX]

@lmeyerov - added two new files and updated per the comments above. Added links from each to the other, and added the system-metrics-to-csv.sh bash script. Please review once more.