Oracle: You wanted SQL on Hadoop? How about SQL on Hadoop and SQL ON NOSQL?
ARE YOU NOT ENTERTAINED?
Oracle wants all technology to speak the same language as the one found in its eponymous database.
Unfortunately for Oracle, the world doesn't work like that, so on Tuesday the company announced new software that lets Oracle aficionados use Oracle SQL commands to probe non-SQL systems, such as Hadoop and NoSQL databases.
The "Oracle Big Data SQL" product means admins don't have to learn other query languages when dealing with non-relational databases or information kept in Hadoop, explained Oracle's big data and analytics president Neil Mendelson in a chat with The Register.
"We can use the Oracle SQL dialect that we've been used to using now - the full SQL dialect - and we can now access directly any one of those three data sources or join any combination therein," he explained.
Similar tools are available in the open-source community, such as Stinger, which makes it possible to use SQL commands to query data kept in Hadoop, or languages like CQL (Cassandra Query Language) which aim to implement SQL commands on NoSQL systems.
Oracle's system, by comparison, is proprietary, and at launch only supports Apache Hive and the Hadoop File System. Documents seen by El Reg say the system has been built to be extendable, so should be able to eventually deal with data kept in MongoDB, Hbase, Oracle NoSQL DB, and so on.
Oracle is "trying to make these new technologies a first-class citizen, embrace them, and allow people to operationalize them in a way that doesn't require them to get new skills or diminish the governance or control that is required from a statutory or regulatory perspective," Mendelson said.
The key tech beneath "Oracle Big Data SQL" is "Hadoop Smart Scan". This tech provides "data-local scanning: data is read and processed at the point of storage," along with "predicate evaluation and projection: only relevant data is transmitted from Hadoop," and "complex parsing: data such as JSON and XML are processed at the source," Oracle explained in a document seen by El Reg.
The Oracle Big Data SQL product will depend upon release 12.102 of Oracle's flagship database. "At the initial release it'll be on Exadata [appliance] on 12.102, then we'll make it available on other engineered machines, then on non-engineered machines as well from a SQL headnode," explained Mendelson. However, "the backend side of it, which is the SQL, is at this point exclusively on our big data appliance."
Oracle may be embracing other types of data, but it's going to do so from the comfort of its high-margin, sticky engineered systems, it seems. ®