OpsDataStore Inc. has updated its statistics-grounded service management suite with improved root-cause analysis and better support for Intel-based servers via integration with the chip maker’s Data Center Manager power monitoring system.

OpsDataStore Uses a dynamic object model to automatically establish relationships between the items in the streams of metrics as the data is ingested (above). For Version 1.2 of its namesake product, OpsDataStore wrote a graph processing engine using a combination of Apache spark and Cassandra to analyze very large numbers of relationships for root-cause analysis. The company built its own engine after learning that existing graph databases were unable to handle the volume multi-node processes its analytics demanded, said founder and Chief Executive Bernd Harzog.

“To best of my knowledge, no one has had to determine the topology of processes that are that complicated,” he said. “We’re talking 100,000 virtual servers, 15,000 applications, 45,000 data stores, 10,000 hosts and an average of 50 transactions per process.”

The result is that OpsDataStore 1.2 can better tie transactions to their impact on infrastructure performance. “Everybody else relies upon correlation, or the the chance that two thing are related,” Harzog said. “What’s new is our ability to supply statistical root-case analysis combined with deterministic algorithms.”

The addition of the Intel DCM metrics to existing utilization, performance and throughput metrics from other vendors such as VMware Inc., AppDynamics Inc., Dynatrace LLC and ExtraHop Networks Inc. enables customers to reduce operating costs by better managing power consumption. DCM basically supplies power-consumption information directly from the server’s central processing unit.

“Over a three-year period you spend more money powering the server than you did to buy it,” Harzog said. “If we can tell people here are your least-utilized servers and the ones that are costing you the most to run, you can consolidate them out of your environment.” Customers can also now calculate operating expense per virtual machine and per transaction, the company said.

Version 1.2 of OpsDataStore also enables ExtraHop customers to integrate wire data into response time metrics for VMware vSphere environments in order to manage capacity and performance in hybrid cloud environments. ExtraHop is a stream analytics platform that turns unstructured network data into structured wire data for analytics processing. “We can now know that a transaction runs on a particular VM and that a throughput metric for a particularly IP address is tied to that VM,” Harzog said. Automatic time-of-day and day-of-week baselines have also been added for every metric.

Pricing is based on the number of operating systems supported at $100 per instance per year.

 Image courtesy of OpsDataStore