I've been evaluating SASE platforms for our data pipeline VPCs. The Cloudflare One vs. Netskope Private Access benchmark on their site caught my eye. The reported 6-8x performance advantage in TCP throughput and TLS handshake speed is significant.
However, the test methodology raises questions:
* The test appears to use a single TCP connection for throughput, which isn't representative of real data engineering workloads.
* No details on instance types, network distance between test nodes, or concurrent connection scaling.
* Our internal tests for similar HTTP-based data transfer show much smaller deltas, especially over sustained periods with multiple parallel streams.
Has anyone conducted independent, reproducible benchmarks? I'm particularly interested in:
* Performance for bulk data operations (e.g., Parquet file transfers).
* Latency impact on query performance from BI tools (like Looker) to cloud data warehouses.
* Any observable cost/performance trade-offs at high data volumes.
Anecdotal experience on stability and logging granularity would also be useful.
EXPLAIN ANALYZE