'

Coolan looks to crowdsource data center analytics

Former Facebook data center and Open Compute Project leaders take their skills to the new start-up.

Hoping to make use of the surprising willingness of large scale data center operators to share the analytics that they accumulate while optimizing their operations, Coolan looks to use this community sourced information as the basis for tools that will allow better data center management through predictive analysis and benchmarking.

fbpv.jpg
Facebook's Pineville data center / credit: Facebook

Coolan believes that proper analysis of the aggregate metrics and correctly applied machine learning can be used to provide accurate predictions of potential data center failures and outages. Coolan's founders, former Facebookers Amir Michael, who was deeply involved with Open Compute, and Jonathan Heiliger, former VP of Infrastructure and Technical Operations believe that the move to white box hardware in the data center gives them the opening necessary for the tools to manage such an infrastructure.

Much like current systems management applications, Coolan deploys code to the servers that it is managing that track a fairly traditional set of server metrics. The difference is that the data from each managed server is compared to the metrics that are gleaned. What will make Coolan different from its competitors is that the analysis of the derived information isn't compared simply against existing server heuristics or a database of standard performance from the server vendor, but rather it is compared against the information that has been derived from all of the participating customers who are providing open analytical data about data center and server performance.

With this information Coolan believes that customers will be able to take a less costly and more proactive approach to data center management by limiting downtime and optimizing hardware usage. Hardware can be replaced or upgraded based on the demonstrable needs of the data center, rather than on a fixed schedule of equipment investment that might not be connected to the realities of day to day operation.