(click image for larger view)
By document-oriented, he doesn’t mean the database stores only documents. On the contrary, it’s more likely storing groups of software objects that might not appear related, but are. When a document is broken down into different pieces of text and images, they are stored in such a system, but many other unstructured data objects find their way into MongoDB as well, say all the information collected on a customer during one visit to the company’s Web site.
Merriman freely concedes that MongoDB and other NoSQL systems, such as Cassandra or CouchDB, are not necessarily good at capturing long-lived, complex transactions. Relational databases are still good for that. But MongoDB can rapidly scale out to capture really large data sets, sort them, report on them, analyze them.
“They’re good at archival. They’re good at event logging,” he noted, then performing analysis on the large data sets to learn things from a disparate set of events.
And they’re catching on fast. “Look at this event,” he noted, “gesturing to the halls filling with MongoDB developers at the MongoSV for Silicon Valley event held Dec. 3. His last West Coast conference was at the former Federal Reserve Building in San Francisco’s financial district, a much smaller venue, and about 200 showed. Now he was hosting 500 attendees in a building meant to stage developer conferences, with video piped from room to room. “This facility is great,” he said, and he thinks the future for NoSQL systems looks bright as well.