6x faster? 🏎 Nextcloud Hub 26 Winter: Performance updates with the biggest architecture upgrade yet

Summary — Nextcloud Hub 26 “ADA” performance update

Key architectural concept

• ADA = Accelerated Direct Access A major rewrite of the database and file-access architecture designed to make Nextcloud scale better and respond faster in highly integrated deployments.

Three main technical changes in the ADA engine

1. Reorganized database access and identifiers

   • Tables and identifiers were split/restructured to allow more direct access patterns.
   • Introduces globally-unique / snowflake-style IDs and creates a foundation for advanced sharding (for example, distributing DB nodes across data centers).

2. Authoritative mount table (replaces mount cache)

   • The previous mount cache (who has access to what) is replaced by a single authoritative table that is updated when shares change.
   • This shifts work from frequent read-time checks during navigation to less-frequent write-time updates, enabling “lean data access” (only fetch the exact metadata needed).

3. Direct S3 file/preview downloads

   • On S3-backed storage, clients can download files and previews directly from S3, bypassing the application server to save network/hardware overhead and reduce latency.

High-performance backends & push model

• New “high performance” backends:
  • Files backend written in Rust.
  • Talk backend written in Go.
• These backends now push state/notifications (file changes, chat events, user presence) directly to clients rather than only notifying. This reduces server load substantially — up to about an 80% reduction in some scenarios.

Assistant and other frontend/backend optimizations

• Assistant / local AI tasks (text/image generation, speech↔text, contextual chat): up to 10× faster.
• Local translation: ~8× faster.
• Mobile background jobs: sped up.
• Previews and photos: ~40–60% faster.
• Talk (fetching/sharing data): ~20% faster.
• UI JavaScript: reduced by ~10% through removal of legacy libraries.

Real-world impact — example

• MagentaCLOUD (Deutsche Telekom) collaboration:
  • Database read/write split reduced load on the heaviest DB node by ~20–30%.
  • End-to-end example: a 10 MB file download dropped from ~2.5 s to ~0.4 s (≈6× faster) on their system.
• Many improvements come from community and partner contributions (including Deutsche Telekom). Some client-side UI changes (iOS/Android photo view) will appear in upcoming app updates.

Roadmap / future directions

• ADA provides a base to push more operations to direct DB/file access and to implement more advanced sharding and scaling capabilities over time.

Guides / further reading

• The presentation repeatedly refers to a detailed Nextcloud blog post describing the ADA engine technical details — recommended for in-depth architecture and implementation notes.
• Expect forthcoming app updates (iOS/Android) and continued community-contributed improvements.

Main speakers / sources

• Nextcloud team / presenter (speaker not named in the subtitles).
• Nextcloud blog post describing the ADA engine (primary technical source).
• Deutsche Telekom / MagentaCLOUD (example partner and contributor).