We've been running a CyberArk Privileged Session Manager (PSM) setup for a few quarters now to handle database admin access. The security team is happy, but my ops team is fielding complaints from the DBA group about latency, especially during peak maintenance windows.
I'm trying to quantify the actual performance overhead introduced by the PSM proxy for high-throughput SQL traffic (think bulk data exports, index rebuilds, large query results). Our setup connects through a PSM component to on-prem SQL Server instances.
Has anyone done similar benchmarking or have real-world metrics to share? I'm particularly looking for data on:
* **Network latency add:** The milliseconds added per query by the proxy hop.
* **Throughput impact:** Any reduction in rows/sec or data transfer rates for large result sets.
* **Connection overhead:** How connection pooling behaves (or doesn't) through PSM.
* **Resource thresholds:** At what point does the PSM server itself become a bottleneck (concurrent sessions, bandwidth)?
I've done some basic `sqlcmd` timing tests with and without the PSM route, and I'm seeing a consistent 15-30ms additional latency per short query, which adds up. For large, multi-GB result sets, the transfer time can be 20% slower. I'm trying to figure out how much of this is our specific network/PSM server config versus inherent overhead.
Any insights on configuration tweaks that helped you (PSM server sizing, tuning, protocol settings) would be hugely helpful. I need to build a case for either adjusting expectations, allocating more resources to the PSM infrastructure, or pushing back on certain use cases.