Recording: RavenDB's Upcoming Optimizations Deep Dive

May 29 2025

RecordingRavenDB's Upcoming Optimizations Deep Dive

time to read 1 min | 60 words

Yesterday I gave a live talk about some of the re-design we did to the internals of RavenDB’s storage engine (Voron). I think it went pretty well, and the record is here.

Would love to hear your feedback!

Tweet Share Share 2 comments

Tags:

More posts in "Recording" series:

(29 May 2025) RavenDB's Upcoming Optimizations Deep Dive
(30 Apr 2025) Practical AI Integration with RavenDB
(19 Jun 2024) Building a Database Engine in C# & .NET
(05 Mar 2024) Technology & Friends - Oren Eini on the Corax Search Engine
(15 Jan 2024) S06E09 - From Code Generation to Revolutionary RavenDB
(02 Jan 2024) .NET Rocks Data Sharding with Oren Eini
(01 Jan 2024) .NET Core podcast on RavenDB, performance and .NET
(28 Aug 2023) RavenDB and High Performance with Oren Eini
(17 Feb 2023) RavenDB Usage Patterns
(12 Dec 2022) Software architecture with Oren Eini
(17 Nov 2022) RavenDB in a Distributed Cloud Environment
(25 Jul 2022) Build your own database at Cloud Lunch & Learn
(15 Jul 2022) Non relational data modeling & Database engine internals
(11 Apr 2022) Clean Architecture with RavenDB
(14 Mar 2022) Database Security in a Hostile World
(02 Mar 2022) RavenDB–a really boring database

Comments

29 May 2025
16:45 PM

Scooletz

What a massive presentation! As a person who spent some time with a db written in .NET I can strongly relate to some points. Personally, I found the following important/worth to mention:

The design of the new pager with atomic swap and lock. Simple, effective, very nice.
Page fault with mmap in .NET and GC. I got hit by it terribly while working with data >> RAM. I think it's the first time I heard it someone mentioning this issue.
File coherence with unbuffered IO. Sorry to hear that you spent some time debugging it. Yes, if you go with unbuffered, some juggling when using mmap is needed.
Stack trace in Raven Studio. This is awesome. I was not aware of the advancements in this area of the product
Write benchmarks. It would be nice to see not only concurrent but also the queue depth / ring IO benchmarked if possible
Shared journal architecture. I read the idea of linking before. Still thinking that it's great
fsyncing for meta on the parent always halts my mind for a second. If fsyncs on files were not enough!

A great presentation with an awesome summary of the performance improvements. Congrats!

03 Jun 2025
08:32 AM

Oren Eini

Scooletz,

Page faults when working with data that is greater than RAM is not an uncommon issue for us. One of the reasons that people use RavenDB is often specifically for that, and it is something that we had to deal with . Fsync on metadata is really annoying, because this is basically the file system not doing its job. That is mostly related to the difference that the operating systems deal with files (small files & frequent operations are far more common on Linux).

Oren Eini

Oren Eini

CEO of RavenDB