Changing over from a legacy system can be hard. It can be harder when you’re in a traditionally brick and mortar industry and have only a few months to do it in.

Kurt Zimmer, the chief technology officer of Room Key, was brought in by a number of large hotel chains such as Hilton, Hyatt, Marriott, and others to create a Kayak competitor. These chains all used old, legacy systems to control booking and recommendations that took way too long to update. Then came almost companies like Kayak and Orbitz that take big fees for each booking but could quickly process data from all over the Internet and return relevant results to users in seconds.

“[Hilton and Marriott are] old legacy systems. They’re slow to deliver. Releases can take months, if not parts of years to get out,” said Zimmer at VentureBeat’s CloudBeat conference today.

So RoomKey was born, and Zimmer was faced with a huge challenge: bring these hotels into that industry in only a few months while convincing the hotel chief information officers that they could trust what he was doing. At the time, the CIOs feared the cloud, Zimmer said.

He turned to Clojure, a “general-purpose” coding language that allowed him to build a scalable, heavy-load bearing recommendation engine to compete with those big aggregator names.

“[CIOs asked me:] What’s your opinion of the cloud? I said, ‘What a joke.’ Everybody nodded, they loved me,” said Zimmer. “It took me six months to spell clojure.”

He worked with Relevance, which builds “tools and platforms that help people take advantage of the cloud.” Zimmer used its Datomic product to build the scalable database, which pulls in information from both legacy systems and the cloud.

Now, Room Key gets a million visitors a day and 32 million interactions with the hotels a day. Four billion pieces of information are processed a day, which he says is three times as much as Kayak. The whole thing costs about $22,000 a month, and he’s experienced zero downtime (hope that’s not a jinx).

Despite all this, Room Key still has some lag issues when it comes to working with those legacy products.

“When you’re dealing with real-time interactions with a bunch of legacy systems, it’s not you that’s the problem,” he said. “We have interesting challenges in that some of our partners frankly can’t respond fast enough.”