I attended a meetup last night in MountainView, sponsored by Oracle.
David Rubin, head of Oracle’s NoSQL team presented the recent advances made to their NoSQL product. This code base was acquired by Oracle 6 years ago from Sleepycat Software, the custodian of Berkeley DB (BDB). Mike Olson, who founded Cloudera and currently is the chief strategy officer (formerly CEO), wrote many parts of the code for BDB during his graduate school days at UC Berkeley under the guidance of Professor Mike Stonebraker. Oracle’s new slogan says – Oracle now offers an enterprise grade NoSQL solution at the most compelling price point in the industry! It seems they are making an aggressive push of this NoSQL product lately. It is based on key-value record abstraction, but many operational features including SQL support have been added to make it enterprise-ready.
Then followed the interesting talk by Mike Olson, a true software engineer with a rich pedigree. He gave a historical account of data management, starting with punch-card based records in the 1960s. The ISAM movement took the record-oriented data processing by storm, as one could re-arrange data in many ways via indexes. The key-value store was abstracted and suddenly files could be exchanged via Unix OS. Much work happened during the 1970s at Berkeley in both BSD Unix and KV store called Berkeley DB. From the wikipedia we find this fact – The founders of the company were spouses Margo Seltzer and Keith Bostic, who are also the original authors of Berkeley DB. Another original author, Michael Olson, was the President and CEO of Sleepycat that provided commercial support for Berkeley DB. Oracle finally acquired Sleepycat in 2007.
He then described how relational DBMS came about and became very popular during the 1980s. But nothing much happened in the database industry for two decades, until recently with many NoSQL products plus Hadoop etc. He compared the current NoSQL movement to what happened during the late 1990s when MySQL came out to support the backend-storage for Internet startups. It had many weaknesses but was adequate as a read-heavy database that was open source and free. There are too many NoSQL databases now, and many will disappear over time. The current funding of MongoDB at a valuation of $1.2B suggests there is perceived value and something is happening.
He also said that platforms like databases are interesting, but the real touchpoint for the customers will be applications such as analytics and machine-learning. He pointed to his iPhone and said in five years that will be your personal assistant. The need for analytics and such software is key to the success of the NoSQL products.
It was a great talk, specially for those of us who spent decades in the database business and I could relate to everything Mike said.