Show HN: S3 compatible store with 1M IOPS(4K-R,p99~5ms), BYOC in 5min with rust
Comments
fractalbits
alexhnn
How does it compare to rustfs ? I saw it claims 1580K IOPS: https://zhuanlan.zhihu.com/p/1977752657562182612
thomas_fa
well I checked the code in their repo，but could not find any snippets as mentioned in the post，which made the claimed perf numbers very suspicious. There was also no cluster configration details in their claim，but you can verify our performance numbers by simply following the deploy steps in the readme.
everfrustrated
>Atomic
>Directories
If you want a file system just use one. Object stores and file systems are different concepts and technologies. AWS EFS exists for a reason.
fractalbits
Yes, people usually consider "S3 is files, but not a filesystem [1], and complaining about the missing features comparing to filesystem. But nowadays, the line between them is becoming blurred. S3 express one zone explicitly called their bucket as "directory bucket", and try to support atomic rename (even only for single object for now) [2]. GCP also has the same move ([3]). Posix fs APIs have its own pros and cons, and we are also considering to extend our metadata engine to support that in the future if it is needed.
[1] https://news.ycombinator.com/item?id=39656657
[2] https://aws.amazon.com/about-aws/whats-new/2025/06/amazon-s3...
[3] https://cloud.google.com/blog/products/storage-data-transfer...
fellowniusmonk
Does this support the the range header?
fractalbits
Yes, you can check the code here https://github.com/fractalbits-labs/fractalbits-main/blob/ma.... For each type of requests we have a separate handler for that.
The basic tests are here: https://github.com/fractalbits-labs/fractalbits-main/blob/ma.... We also have fuzz tests internally, which could be open once we feel it is ready.
We wrote a blog explaining why we built another object storage and why it is different: https://fractalbits.com/blog/why-we-built-another-object-sto.... Welcome any useful feedback!