Is the database still relevant? Panelists at a Software Development Forum session in Palo Alto, debated this and other questions about the role of databases in the future, discussing the elevated role of XML, commoditisation, and open-source software.
While considering whether the database is merely becoming a component of part of a larger application platform, MySQL AB co-founder David Axmark said he would never put data in a format that he did not know.
"I would like to have data in a way I could handle," he said.
IBM's Jon Rubin, senior product specialist for DB2 solution development, said a business executive might want to purchase a whole platform in which the particular database is irrelevant to him.
"But if you're IT , that database is not a commodity, not today, and probably not interchangeable," he added, citing issues such as manageability, recoverability, and service-level issues related to specific databases.
Kevin Perry, chief data architect at Inovant, which provides credit card transaction processing services for Visa, said the role of the database depends on the depth of the application stack.
Light versions of DB2 or Oracle databases perhaps could be displaced by Java Data Objects. "My problem deals with terabytes of information, so the value for us is data itself," Perry said.
"I think the database should be a commodity and I shouldn't drill down so much in Oracle's calls or (IBM DB2) Universal Database's calls or anyone else's calls," he added. But drivers should be swappable for putting in, for example, DB2 Universal Database in place of another database, Perry said. There are times, however, when a specific feature of a particular database is desirable, but users need to have flexibility on price.
IBM's Rubin stressed the complexity of enterprise databases makes specific variants relevant.
"In the real world of enterprises, the database is so complex and so challenging in so many different directions, it's hard to find an application that can afford to live with the lowest common denominator of, let's say, ANSI SQL," he said.
Although panelists noted the upcoming release of the SQL 2003 specification, they also said that not all features of the SQL 99 specification have been put to use.
"If you're familiar with the SQL 99 spec, there's a vast body in SQL 99 that's still untapped," Rubin said.
Meanwhile, more functionality is being expressed in XML, he said. "If we can make XML a first class data type, what that means is we would actually optimise access to XML."
Axmark added that XML is useful as an exchange format.
Rubin also said unstructured data, such as digital images and digital audio, do not belong in databases, but meta data references describing these data sets should be in the database.
Panelists also discussed the role of open-source databases.
"I would say we do some things new," Axmark said. MySQL offers an open-source database.
"We do our innovations at a very, very low level," Axmark said.
Rubin called the open source database "an important phenomenon."
"I don't see it being about ease of use because there was already pressure on database suppliers because of Microsoft," Rubin said. Open source is bringing pressure on the cost model and is having a profound impact resulting in the commoditisation of Intel hardware, he said.
Perry, however, expressed reservations with the level of support available for open-source databases. "I need to be able to put my hands on someone's neck and say, 'This doesn't work,'" he said.
"I pay a large fee to Oracle or IBM to say, 'This is broke.' My only fear with open source is this: I don't have that same neck to grab."
But Axmark responded that MySQL has support round the clock.
An audience member charged that databases would be commoditised similar to the way Linux, Apache, and Eclipse have prompted commoditisation via open source. "Why isn't it going to happen in the database? It just seems inevitable," the audience member said.
Eyeing the future of databases, Perry cited the concept of the DAN, or Database Area Network, in which the database would, through virtualisation, pull compute cycles from systems.
Rubin said databases would not consume all data in the enterprise. "It will get more of the data. I don't think it will ever get all the data," he said.
He also stressed the need for more flexible licensing models for databases, such as by offering a per-transaction licensing option. "The commercial databases are either per-processor or per-user. That's not enough flexibility," Rubin said.
Paul Krill writes for InfoWorld