Effective performance management relies on having access to the right information on a timely basis. The more data points on a performance indicator you collect, the more complete — and, hence, the more accurate — your view will be. The more important the network link, the more accurate your data needs to be. For years, five-minute polling has been the industry standard, but that frequency is woefully inadequate for a critical link that carries latency-sensitive data.
If, for example, the polling interval is set at 1-minute, a 1-second spike in bandwidth utilization will be completely flattened and impossible to diagnose because 1-second represents only 1.7% of the total monitoring time. In the typical 5-minute poll cycle of most monitoring systems, those critical seconds will appear as little more than 0.33% of the total elapsed time — well below the radar of most network managers. Nevertheless, this 1-second spike is of sufficient duration to disrupt automated trading applications, online financial transactions, and other latency-sensitive communications.
While the SevOne Data Platform can be configured to support polling high frequency polling, new streaming data sources on key network devices from Cisco, Juniper and more are now available. Now, instead of polling devices, key performance metrics can be streamed directly to SevOne Data Cloud, where real-time analytic policies set by users run continuously against the real-time streams at ingestion - including, aggregations, baselines and thresholds. This capability affords a competitive edge to time-sensitive business processes, such as high-frequency trading in the financial industry.