NoSQL systems are still in an early phase, but already they've played a key role in the development of massive social networking sites Twitter and Facebook as well as social gaming site Farmville. Their speed and reliability cater to a Web world, an appeal that will only grow with the mobile Web.
They don't pretend to do everything a relational database does, and a mere 5% of companies are using or piloting them, our survey of 755 business technology pros finds. Heck, 44% of survey respondents hadn't even heard of NoSQL. But we think it's inevitable that NoSQL will play a role alongside conventional database systems in many enterprises.
In many cases, NoSQL systems are bringing data management techniques into the enterprise that originated to meet the needs of Web businesses such as Google, eBay, and Amazon, which were generating data at rates that choked relational databases.
Not since Edgar Codd, a database pioneer at IBM, issued his "12 rules" of relational databases in the 1980s has the market been in such ferment. Startups are popping up to commercialize the open source NoSQL projects, including Cassandra, MongoDB, and CouchDB. Suppliers of the tools surrounding conventional database systems are getting in on the act as well. For example, Quest Software's Toad, typically used to automate Oracle, IBM DB2, and Microsoft SQL Server administrative tasks, is now available for Cassandra, an open source project that began at Facebook. Embarcadero Technologies is extending its relational tools to support NoSQL systems. While enterprise use is tiny today, 22% of the IT pros we surveyed say they're interested but need to learn more.
The appeal of NoSQL is that it handles mass quantities of data, quickly, across a cluster of servers that share resources, making it both fast and reliable. The fact that it's open source keeps costs down, and it's easier to use than conventional databases.
Once Eric Evans, while a systems developer at Rackspace, used the term NoSQL to describe large, clustered but nonrelational data management systems, this trend had the provocative name it needed to take on "movement" proportions. Projects that use NoSQL approaches include Membase Server, Scalaris, LightCloud, and RavenDB. Pincaster applies the NoSQL approach to a data system for applications using geographical data.
Make no mistake, NoSQL systems aren't as good as relational databases at many things. Relational systems, with their strictly defined properties, can start a multistep transaction and guarantee the integrity of its data upon completion--no mean feat in a world of rapidly changing data. Relational databases pass the ACID test--atomicity, consistency, isolation, and durability of transactions. NoSQL can't.
But VoltDB, the latest startup from Michael Stonebraker, points to where these two worlds may meet. VoltDB is a relational database that handles millions of transactions a second and still meets the ACID test by distributing both the database and the data across a server cluster. That's similar to how the NoSQL Cassandra, Redis, and MongoDB distribute themselves to manage high volumes of unstructured data.
Download the September 20, 2010 issue of InformationWeek