First class caching
First class caching
Reza Naghibi - August 5th 2026
I will be the first to humbly admit that FiberFS has an advantage over all other filesystems. It has the advantage of being built from scratch to solve a very specific problem: how can you run a POSIX filesystem on S3. You might be asking, why S3? Well, the future won’t be running on your locally attached disk. I’ve been told the future will be big, and will require vast amounts of compute, memory, and storage resources. Enter FiberFS.
Running a POSIX filesystem on S3 is no easy task. In most cases, you have to make a choice, do you want speed or do you want correctness? FiberFS brings balance to this equation, it offers both speed and correctness, and with a good measure of simplicity on top (did you know FiberFS does not require any kind of database?). It does this with a very intelligent core design and, most importantly, everything FiberFS produces is cacheable.
Caching is the key to good performance. Infact, I will even say that you won’t achieve the highest levels of performance without caching. You might say to yourself, I don’t use any caching and everything works fine for me. Well, that's not necessarily true. Pretty much every system or component you interact with likely has caching built into it. Your CPU has a cache, your disks have caches, network cards and GPUs have caches, even memory has caches. Databases have caches, operating systems have lots of caches, and of course there is the internet, its chock full of caches. There is even a global multi billion dollar caching industry and these caching companies are called CDNs. So whether you know it or not, a lot of what you are doing is made possible by caching.
And this is the beauty of FiberFS, by simply using it, you now have access to one of the most cacheable filesystems available. Anything can be cached anywhere. Want to store your filesystem on a CDN, no problem. And yes, FiberFS comes with its own intelligent built-in cache. This cache will keep the files and directories you use local while efficiently storing everything on S3 using 100% correct POSIX semantics. What this means is that using FiberFS will look and feel like you are using your local disk, but it will have the resilience and scale of S3.
Also, FiberFS’s cache can be clustered. This means if you are running FiberFS across multiple hosts, FiberFS is smart enough to distribute and share its cache across all of those instances. This acts as a performance multiplier, it gives you more cache and even better performance. Normally when you hear the word cluster, two things come to mind, expensive and complicated. FiberFS reverses this, clustering is baked into the product at no additional cost and it's simple to setup and requires little to no maintenance and support. Does this mean that FiberFS can be used in HPC setups? Yes, it does.
So what should you do next? Give FiberFS a spin. Getting started docs are here. And don’t forget to let us know what you think.
Blog
Home