Sizing
Sizing your TP instances largely depends on your workload, and, when possible, we recommend testing your TP workload to determine your actual numbers. In general, though, most TP workloads are CPU-intensive and benefit from the same guidance.
Estimations
The following numbers are starting points for estimating your instance sizes. They’re based on our experience working with TP users but will vary depending on your workload. Its important to note that these numbers depend on the number of transform functions applied to the data.
Measurement | Size | Throughput |
|---|---|---|
Unstructured log event | ~256 bytes | ~10 MiB/s/vCPU * |
Structured log event | ~768 bytes | ~25 MiB/s/vCPU * |
Metric event | ~256 bytes | ~25 MiB/s/vCPU * |
Trace span event | ~1 KB | ~25 MiB/s/vCPU * |
* These numbers are conservative for estimation purposes.
* - 1 vCPU = 1 ARM physical CPU or 0.5 Intel physical CPU with hyperthreading.
Due to Vector’s affine type system, memory is rarely constrained due to data processing. Therefore, we recommend 2 GiB of memory per vCPU as a general starting point.
Cloud | Recommendations |
|---|---|
AWS | Latest generation Graviton (recommended) or Intel Xeon, ≥8 vCPUs |
Azure | Latest generation Intel Xeon, ≥8 vCPUs |
GCP | Latest generation Intel Xeon, ≥8 vCPUs |