Joyent Offers NoSQL Database As A ServiceJoyent Offers NoSQL Database As A Service

9 More Cloud Computing Pioneers

A Joyent data center in Northern Virginia will use the the Cloudant NoSQL database system on a server cluster to offer its first database as a multitenant service.

While database software such as Oracle or MySQL has been available for some time on Amazon Web Services, only a handful of cases exist where a customer may go to a service provider and get a scalable, NoSQL database as a service. A predecessor service meeting that description is Amazon Web Services' DynamoDB, which is still a beta offering.

Joyent is a San Francisco infrastructure-as-a-service provider that has marched to a different drumbeat from its entry into the cloud market. The firm's staff includes Sun Solaris expertise and it builds infrastructure based on an open source derivative of Solaris that it calls SmartOS. Gartner has put Joyent in its "challenger's" quadrant for the cloud marketplace.

By bringing a third-party NoSQL system to its cloud, Joyent is one of the few service providers able to offer a highly scalable data management system as part of its service offering. Joyent already has a widely instrumented environment that could be useful in maintaining database operations. It makes use of Dtrace, a utility within SmartOS for finding application or system problems in near real time.

[ Want to learn more about the "different drummer" characteristics of Joyent's cloud service? See Joyent's Cloud Competes With Google, Amazon. ]

It also operates a leaner form of virtualization than typical x86 environments based on Solaris-like "zones" or "containers." One SmartOS host can run many virtual machines under the host operating system instead of each VM needing its own, as is the case in the VMware, Citrix Systems and Microsoft Hyper-V worlds. The result is a leaner, virtualized environment with less overhead, one that comes closer to the bare-metal environment in which applications run best, said Jason Hoffman, Joyent founder and CTO, in an interview.

"With their virtualization layer, we get all the performance of bare-metal servers," said Alan Hoffman (no relation to Jason Hoffman), co-founder and chief product officer of Cloudant, in an interview.

Cloudant is an independent software maker that builds a scale-out NoSQL system for multitenant use out of components from the Apache CouchDB open source project. Some version of the Cloudant system is running under several service providers, including SoftLayer and Rackspace, but Cloudant has chosen to launch what it calls Cloudant Database as a Service (DBaaS) on Joyent. The first site is on a cluster in Joyent's Ashburn, Va., data center, and it plans to roll out DBaaS in all seven other Joyent locations.

"We liked the durability of the data model" in CouchDB, explained Alan Hoffman. CouchDB can recover from the loss of one or more copies of a data set, and Cloudant has built "a more scalable, robust service on top of that."

That may sound boastful, but Alan Hoffman said it means a user of its DBaaS will be assured in the near future that his data will survive "even if the entire data center is lost." A single Cloudant system will be able to be set up to run with part of its operations in a second data center, he said.

Joyent meters the system's use and charges the user. Early adopters and test drivers won't be penalized for trying Cloudant DBaaS. Any charge less than $5 will be forgiven, said Jason Hoffman. DBaaS is billed by the hour, with number of HTTP requests and the nature of queries -- light or heavy -- determining final charges. Customers may opt for a dedicated Cloudant DBaaS, which will be billed on a per node of use basis, Alan Hoffman said.

Cloudant is a five-year-old software maker in Boston, founded by three MIT particle physicists who were looking for a distributed big data handling system. It has 45 employees, Alan Hoffman said. It is backed by venture capital from Avalon Ventures, In-Q-Tel, Y Combinator and Samsung Venture Investment. Samsung invested in the firm because its distributed data management and capture system is believed to have potential uses in future mobile and Web applications, Alan Hoffman said.