Abstract

The datacenter of the future will rely on artificial intelligence (AI) to drive predictive & proactive management of the server infrastructure. In turn, AI systems rely on ingesting massive amounts of granular data. iDRAC9’s Telemetry Streaming, available with the new iDRAC9 Datacenter license, is architected to provide highly granular, high-value data using resource-efficient data streaming.

Dell EMC commissioned Tolly to benchmark the number of data points collected from a representative PowerEdge server via iDRAC9 v4.00.00.00 in a 24-hour period and compare that to a prior, less data-rich monitoring implementation. Additionally, data collection via polling and streaming were compared for network efficiency.

Tolly found that iDRAC9 Telemetry Streaming running on a representative test system produced over 2.9M data points in a 24-hour period compared with 6,240 data points for a typical polled server using the existing Redfish interfaces and a reasonable real-world polling interval of 30 minutes. With Telemetry Streaming each 24-hour report can be collected via a single HTTP request compared with a per-report best case of 17,280 HTTP requests to collect a single report daily.