MongoDB, the company behind the document non-relational database of the same name, has released its 5.1 update but only as a managed service.
While critics voice concerns that bugs from 5.0 have still not been fixed, the NoSQL poster child has pushed on with efforts to accommodate time series data and better support analytics on operational systems.
The result could be good for users already requiring these features, but is unlikely to usurp rival databases with these performance specialisms.
MongoDB 5.1 is the first in an "accelerated release cadence" and will only be fully supported on MongoDB Atlas, the company's database as a service, while being also available to download as a "development release."
It features support for globally distributed time series applications through MongoDB's native sharding to horizontally distribute massive data sets and co-locate nodes.
Among other features of note is improved support for analytical applications by allowing
$graphlookup functions to combine and analyse data distributed across shards, something which was not previously possible, according to a company blog.
CTO Mark Porter told The Register that the inability to have joins across sharded collections – groups of MongoDB documents – had been "a real adoption blocker for our customers."
"What people want now are things like fraud verdicts and cart recommendations; they want to do all those other things in real time in their app," Porter said.
Previously, developers were only able to perform query joins from a sharded collection to a non-sharded collection. They couldn't use
$lookup do a join between sharded collections, he said.
"Sharding is such a fundamental piece of scaling out MongoDB; now that they can do this, it just removes that friction," Porter said.
But time series and analytics capabilities were not at the forefront of concerns among those using the database, according to one commentator.
Kimberly Wilkins, MongoDB technical lead at open source support and services company Percona, said release stability was a much greater concern among the developer base.
She pointed out that MongoDB was only providing one major release per year for on-premises and via the community edition, "with all other dot point releases going only to their customers that are using Atlas."
The versions following 5.0 "have been problematic for users so far," she claimed, with bugs impacting through to the release of MongoDB v5.0.3 on September 21, 2021.
Those first three releases were all labelled with the warning: "MongoDB version 5.0.0 is not recommended for production use due to critical issues..." The bugs caused issues such as duplicate unique keys, omitting a page of data, data loss, and problems restarting.
"With v5.0.x being so buggy, I was hoping that MongoDB might be forced to offer the updates in v5.1.x, v5.2.x, etc, to all their users. However, it appears that they still only plan to offer those between versions to their Atlas customers for production use. This would mean community edition users would not benefit from those fixes until MongoDB releases v6.0, expected sometime around mid-2022," she said.
"From a community standpoint, this would have a significant impact on how developers would experience working with MongoDB – far from encouraging more people to standardise on MongoDB, it would encourage them to look at other options for running their workloads," Wilkins claimed.
- Judge tosses NEC's claim that Oracle salespeople tricked it into using the wrong software license prior to audit
- And just like that, Amazon Web Services forked Elasticsearch, Kibana. Was that part of the plan, Elastic?
Although a MongoDB spokesperson said 5.1 would be available for Atlas, Enterprise and MongoDB Community users, closer examination reveals the second two releases are development releases for "evaluation only".
MongoDB was not aware of any critical issues in 5.0.3 or 5.0.4, a MongoDB spokesperson said.
"Our standard operating procedure is to always test, review and improve our releases on an ongoing basis. Transparency has always been at the forefront of our values. We openly communicate any critical issues within our software and share those publicly as soon as information is available," the spokesperson added.
Matt Aslett, research director at Ventana Research, said to some extent the problems were to be expected with a major release, and the change in cadence "will alleviate some of those things which are hard to avoid."
He said the new time series feature would please developers already using MongoDB for that purpose but would be unlikely to discourage those looking at specialist time-series databases for specific projects.
Similarly, with analytics, the new features would encourage those already doing these kinds of queries within the NoSQL database, rather than impact choices of specialist databases.
"They're not going to be used as a data warehouse – I don't think MongoDB is remotely claiming that to be the case. But for things like quick fraud detection or recommendations that rely on an analytics process, but delivered as an operational application, then this kind of functionality is increasingly what enterprises are looking to deliver," he said.
In terms of MongoDB as an enterprise database option, more companies are seeing it as a "strategic data platform" but there are "many, many more examples of it being used for more tactical applications," he said.
"A lot of that is to do with the functionality and capabilities which evolved over time, while some of that is to do with the sales and marketing approach. Clearly, MongoDB has been very successful in selling to and marketing to developers. They want to retain that, but at the same time, add a focus on [higher] decision-makers," Aslett said. ®