MongoDB - NoSQL database purveyor of the document-store variety - is rolling out a multi-cloud cluster it said will let users deploy a single distributed database across AWS, Google Cloud, and Microsoft Azure.
Atlas, the Database-as-a-service (DBaaS), was first launched in 2016 but has so far been restricted to running on a single cloud provider. MongoDB said its multi-cloud cluster will allow users to boost geographic reach across the 79 cloud regions supported by the top three hyper-scalers. Resilience and compliance could be the benefits, the vendor added.
Mat Keep, senior director at MongoDB, said the widened service is designed to also help those developers wanting to pick-and-mix tools from the respective cloud giants.
“We have a lot of customers who really like us using Lambda, the serverless functions on AWS, but I also like using things like TensorFlow, and the machine learning AI stuff on Google. With multi-cloud clusters, you can have some nodes of your cluster running on AWS, maybe serving your kind of e-commerce application, in the background, MongoDB Atlas can pump data into Google where you can do all of the AI and ML stuff on users browsing habits to create recommendations for example,” he said.
MongoDB wants developers to see its database as a general-purpose system, not just in the applications where the company has built its core markets, such as content management and catalog systems.
Keep said about a third of commercial engagements were now migrations from relational databases, pointing out that adding elements like multi-document ACID-compliant transactions a couple of years ago had "opened up a whole lot of back-office applications, such as payment systems".
For example, European payment provider NETS Group has moved its transaction system to MongoDB, he said.
According to an IBM survey, by next year, 98 per cent of companies plan to use multiple hybrid clouds, so the thinking behind MongoDB creation of a multi-cloud DbaaS may be sound, but it is not the first to come up with the idea.
Andy Pavlo, associate professor of databaseology in the computer science department at Carnegie Mellon University, said: "It's an impressive engineering feat, but it is not a novel idea: other DBMSs have supported it for a while."
He said that users who were “really paranoid about availability” might be attracted to the new offer. It might also appeal to users trying to avoid lock-in to a cloud vendor database. AWS has Amazon DocumentDB, while Azure has Cosmos DB.
As for MongoDB’s claim that its DBaaS is general purpose, it was not quite what it seemed, Pavlo said.
"General purpose is not general purpose. MongoDB is still primarily used for the front end, what I call operational workload: things that are backing websites, things where you're ingesting data, for example," he said.
Although it had added analytical features, "you wouldn't want to replace Snowflake, AWS Redshift or Google’s BigQuery with Mongo," he said.
"There are certain aspects of the architecture that mean it's simply not going to do analytics with the same performance as like the specialised data warehouses," he said
While MongoDB had added transactions, the inertia of customers works in favour of popular systems from companies such as Oracle. For many, the thought of rewriting the semantics of transactions and SQL-queries to go from Oracle to Mongo would most likely be too great an undertaking to consider, said Pavlo. ®