ZDNET’s key takeaways
- In relation to databases, virtually everybody nonetheless makes use of out of date packages.
- Price is a serious purpose firms aren’t upgrading
- Unsupported software program comes with a number of safety points.
All of us use out-of-date software program. Typically we do it out of laziness, generally upgrading prices an excessive amount of, and generally it simply works, and we don’t wish to change it.
Nonetheless, when Percona, the open-source database administration system (DBMS) assist firm, revealed in its 2026 State of Open Supply Database Administration Report that 97% of US-based database directors (DBAs) have been nonetheless working outdated DBMS packages, I raised my eyebrows. Actually!? That’s quite a bit, and DBMSs are mission-critical for all firms.
Why we hold utilizing outdated DBMSs
In keeping with Percona, firms proceed utilizing DBMSs past official end-of-life primarily as a result of upgrading is a high-risk, cross-stack migration challenge, not a easy software program refresh. Legacy software dependencies, deprecated options, shopper compatibility, modified defaults, testing calls for, and the necessity to keep away from manufacturing disruption can all make postponement seem safer or cheaper within the brief time period. In different phrases, they’d quite be secure than sorry.
After all, that presumes they’ll keep forward of safety breaches attributable to persevering with with end-of-life software program. Good luck with that. That’s by no means guess lately, when AI can discover safety holes sooner than we will repair them.
Different causes, in keeping with Percona, embrace that database upgrades are hardly ever easy model adjustments as a result of they’ll expose dependencies all through the applying stack. Legacy functions might depend on deprecated or eliminated database options, requiring groups to determine these dependencies and revise affected code. For instance, MySQL 8.0’s default caching_sha2_password authentication technique could cause bother for older shoppers and should require configuration adjustments.
Additionally: Switzerland takes a Swiss Military knife to Microsoft 365
Different adjustments require cautious application-level overview. A shift in default character units, for instance — from latin1 to utf8mb4 — can not safely be assumed to be clear to present functions or data-handling routines. Percona additionally advises organizations to evaluate the total atmosphere, together with functions, working techniques, libraries, and configuration recordsdata, since a database migration typically impacts way over the database server itself.
That complexity raises the associated fee and threat of shifting to a supported launch. Groups must run pre-upgrade checks, validate the migration in staging, and schedule manufacturing work fastidiously to restrict disruption — duties that require specialist experience, engineering time, and upkeep home windows.
Main launch transitions can add nonetheless extra issue. Percona cites the improve from MySQL 5.7 to eight.0, which introduced a considerable data-dictionary redesign and eliminated the question cache, for example of why “simply improve” isn’t an ample plan.
Additionally: ‘I can’t cease’: 80% of builders discover AI coding extra addictive than useful
Within the broader backdrop, Percona discovered that database groups are constrained by rising cloud prices, efficiency issues, wasted engineering time, downtime issues, scaling complexity, fragmented insurance policies, and a scarcity of expertise. These pressures scale back the organizational capability out there for lifecycle work resembling database upgrades.
Cash, cash, cash
Price pressures have been the survey’s most incessantly cited impediment to database total-cost-of-ownership reductions and upgrades.
The findings put cloud consumption prices (31%) forward of licensing (23%) because the main reported barrier. That issues as a result of shifting to managed database platforms can remove some operational work whereas additionally creating much less predictable payments as storage, compute, community site visitors, excessive availability, and data-transfer wants rise.
“Cloud economics are altering, AI is introducing new workload patterns, laws have gotten extra complicated, and organizations are more and more acutely aware of the place their information and expertise are managed,” Percona CEO Peter Farkas mentioned within the firm’s announcement. “The organizations that can be finest ready should not essentially these that may predict what comes subsequent. They’re those that protect the liberty to adapt when it arrives.”
Efficiency and coverage gaps
The report additionally depicts database operations as an engineering-efficiency and resilience drawback, not merely a query of question latency:
- 42% cited inefficiency or sluggish throughput as a efficiency problem.
- 41% cited wasted engineering time.
- 40% cited downtime.
- 37% cited issue scaling.
On governance, 54% mentioned vendor lock-in makes it tougher to adapt to altering regulatory necessities. Greater than half reported fragmented instruments or insurance policies, whereas 46% recognized each software program supply-chain traceability and a reactive safety posture as issues.
These outcomes hyperlink the legacy-version subject to broader administration challenges. Unsupported software program can complicate vulnerability administration and audit necessities, whereas fragmented controls and database sprawl make upgrades and migration planning tougher to execute persistently.
AI will not be mainly a characteristic hole
For AI readiness, respondents have been extra prone to determine foundational organizational obstacles than a scarcity of database options. Fragmented information structure and expertise shortages tied as the highest obstacles, at 23% apiece.
That result’s a helpful corrective to the present rush so as to add vector search, embedding storage and agent-oriented capabilities to database choices. These capabilities could also be helpful, however Percona’s respondents counsel that disconnected information estates and a scarcity of mandatory employees experience are extra rapid blockers.
Additionally: Open weights vs. closed: An AI civil conflict’s afoot
“Database leaders must construct for change, not for a single model of the longer term,” Farkas mentioned. “Management and suppleness have to be necessities, not afterthoughts.” Nobody is aware of the place the AI revolution will take us. For that purpose, due to open-source flexibility, good firms nonetheless put their DBMS assets into open-source DBMS.
