This article is more than 1 year old
EDB tries to crowbar graph, JSON, and time-series data models into PostgreSQL – but can they pull it off?
Specialist databases will still be necessary, analyst says
EDB, a prominant backer of the PostgreSQL open-source database, expects to focus on graph, JSON, and time-series data in the upcoming autumn release. Analysts, however, are sceptical about its ability to optimise for different data models ahead of built for purpose databases.
Last week, EDB announced a 59 per cent increase in annual recurring revenue, although being privately held it can pick and choose which financial metrics to release. Its team has grown by nearly half, to 300, however that is dwarfed by comparable open-source-supporting firms like Red Hat, with 13,000 employees.
Still, EDB said it sees the database market as up for open-source takeover in the same way Red Hat helped unsettle the status quo in the enterprise Unix market over the last two decades.
Speaking to The Register, EDB CEO Ed Boyajian, former Red Hat veep and general manager of North America, said the company is pushing the community's efforts to strengthen the features supporting JSON and JSON-B, including the alignment of PostgreSQL's JSON capabilities with the SQL standard. Similar work would support network-like graph data in a single database.
He said EDB wanted to help users support these data models without the need to reach for specialist databases like Couchbase or Neo4j, for example.
"The idea is that they can use a common SQL standard, and in a single database we see an increasing appetite among users," he said.
Postgres was first proposed in 1986 [PDF] as a successor to Ingres. PostgreSQL is released under the OSI-approved PostgreSQL Licence. As a relational database with an object-relational data model, it is easy to extend Postgres to work on new data types such as graph and JSON, Boyajian said.
EDB is promising to help build on these features into PostgreSQL 14, expected in the third calendar quarter, to make multiple data models addressable in SQL. The same would apply to graph and time-series data, which is supported by its own family of specialist databases, Boyajian said.
"The multimodal nature of Postgres is really changing the landscape," h told us. "There are time-series, graph, and document database companies, but these are all extensions inside of Postgres and I think that's one of its inherent superpowers."
EDB is not alone in trying to build new data models into a relational database. Last year, Oracle released its 21c database, which Big Red said would offer JSON as a native data type. MariaDB, another open-source relational database, also supports JSON.
Carl Olofson, IDC research vice president for data management software, doubted either approach would provide the whole answer for those wanting to make use of unstructured or graph data part or their enterprise stack.
"Optimising for relational data management is fundamentally different from optimising for other models, such as document or graph, so handling these other models very well requires an alternative parallel data access and management approach, not just supporting non-relational structures in what is basically a tabular scheme," he said.
Although EDB and other PostgreSQL contributors may be addressing these issues in the coming release, admins may still want to look at specialist tools for different data types.
"For very demanding edge-data-handling databases, an application that uses a pure-play JSON database, such as MongoDB or Couchbase, may still be the way to go," Olofson said. "For deep and complex graph analytics, using a native system such as Neo4j or TigerGraph may be called for. If one wishes to retain JSON document and graphs, and use them in combination with relational tables, then a multi-model approach makes sense, and the PostgreSQL option could work out quite well."
Banking software firm tiptoes off to the cloud with MariaDB after $2m Oracle licence shockerREAD MORE
The IDC analyst noted PostgreSQL's recent popularity among both open-source users wanting to avoid dependence on MySQL and those migrating away from proprietary databases.
Last year, German insurance services firm BG-Phoenics moved from IBM Db2 databases to the EDB iteration of Postgres, while Indian credit reference agency TransUnion CIBIL moved from Oracle in the same direction.
But PostgreSQL is not the only open-source database appealing to proprietary users. Financial services software specialist FNI moved to MariaDB to avoid $2m Oracle licensing fees, for example.
Hence EDB's play to be more than just an Oracle or Db2 replacement. Whether or not promises to offer enhancements to tackle multi-model data in a single SQL database come to fruition, users will need to see what EDB and the developer community come up with later this year. ®