One of the gotchas of public cloud infrastructure is penalty charges that can kick in when a user exceeds the amount of computing resource budgeted at startup. The newest player in the emerging database-as-a-service market (DBaaS) in the U.S. thinks it has a better way with a NoSQL service that bills usage after the fact while spreading its database footprint across a theoretically infinite number of servers.

Riga, Latvia-based Clusterpoint, SIA is entering the U.S. with the opening of a data center in Dallas and strategy focused on small and midsize business clients that have unpredictable database loads. It’s headed by Zigmars Raascevskis, a former Google engineer who worked on the database for that company’s search engine.

In much the same way as Google developed technologies to scale databases almost infinitely, Clusterpoint has built a database stack across a combined machine footprint, which it says not only delivers rapid and nearly infinite scale but also better performance. “Even when the total amount of work is the same, running across multiple parallel databases makes each job execute faster,” Raascevskis said. “In the cloud, the developer’s database request can be distributed among hundreds of servers in parallel, and that significantly reduces clock time to get a result.”

Conventional database services in the cloud require users to specify how many resources will be deployed, a requirement that limits flexibility, Raascevskis said. Clusterpoint charges only for resources used rather than provisioning a set amount up front. This avoids the penalty fees that can accrue when developers go beyond their allocated resources. Clusterpoint estimates that its model can reduce infrastructure costs by one-third or more. Pricing is “in same ballpark” as other DBaaS services, Raascevskis said.

The company’s product and pricing are particularly also useful in scenarios in which databases grow quickly, such as aggregating multiple databases for analytics. Given its origins, it’s not surprising that Clusterpoint is also a good search engine database. It’s ACID-compliant and provides transactional consistency, the lack of which is one of the common complaints about the NoSQL model. Raascevskis called the query language a “simplified version of SQL with its own syntax.”

Written in C++, the technology is not available in open source form, but there is a freemium service that provides up to 10 gigabytes of storage at no charge and permits registrants to download an on-premise version, paying only for technical support.