Ex-Facebook engineer launches startup to tackle server management


One of the founders of Facebook’s Open Compute Project has launched a new company that aims to cut the cost of running data centers using “community-based analytics.”

Coolan announced a beta version of its first product on Wednesday, a service that collects and aggregates data about its customers’ server environments and uses it to predict failures, prevent outages and allow companies to benchmark themselves against peers.

The company was cofounded by Amir Michael, a former Facebook engineer who led development of the Open Compute Project’s first server designs. That group aims to give customers more control over how their equipment gets designed and built, and Coolan aims to have a similar empowering effect.

Customers must first install a small piece of software on their servers that collects information about the components installed, how old they are, their error rates, what firmware they’re running and other variables.

Coolan pulls the data into its cloud, and customers access it via a dashboard or programming interface to give them a better picture of what’s happening in their data centers. More importantly, Coolan aggregates the data from all its customers and applies machine learning techniques to uncover patterns and trends.

If a customer has an outage, Michael says, Coolan will be able to quickly pinpoint the likely cause by looking at historical data. So whether a company has 100 servers or 1,000 servers, it will get insights from a much larger data set.

Server vendors offer their own management tools that can predict failures, but Michael says customers are better off with data from a third-party that doesn’t have a vested interest in the results. Coolan will provide comparative data, he says, such as which models of disk drive from which vendors perform most reliably.

And because it’s a cloud service, “our data set will be much larger,” he says.

That will depend on how many customers it can attract, however. Data centers have traditionally been guarded about sharing operational data, but Michael says there’s a trend towards openness as companies struggle to reduce costs.

“We’re riding on the tails of Open Compute,” he says. “People are starting to feel more comfortable about sharing information about their infrastructure.”

He also acknowledged a “chicken and egg” problem. Coolan’s service will be more useful when it has data from a lot of customers, but it has to attract those customers to achieve its potential.

Still, Coolan says its service provides useful insights even in the context of a single customer, because it lets them see what equipment they’re using and how it’s performing.

The service collects the data via a secure HTTPS connection, and information about individual data centers isn’t shared, he said. Customers can see what data is being collected and opt out the parts they don’t want to share.

The service is being tested by a handful of companies, with a wider roll-out to follow soon. Coolan is still figuring out the pricing, including whether it will offer site licenses or per node licenses.

Must read: Hidden Cause of Slow Internet and how to fix it
View Comments
Join the discussion
Be the first to comment on this article. Our Commenting Policies