+972 52-548-6969
Working on Voron…
time to read 1 min | 53 words
This took a bit less than what I expected, but…
And yes, it works. And this is running on Ubuntu.
And no, it isn’t ready.
FUTURE POSTS
- Partial writes, IO_Uring and safety - about one day from now
- Configuration values & Escape hatches - 4 days from now
- What happens when a sparse file allocation fails? - 6 days from now
- NTFS has an emergency stash of disk space - 8 days from now
- Challenge: Giving file system developer ulcer - 11 days from now
And 4 more posts are pending...
There are posts all the way to Feb 17, 2025
RECENT SERIES
-
Challenge
(77):
20 Jan 2025 - What does this code do? -
Answer
(13):
22 Jan 2025 - What does this code do? -
Production post-mortem
(2):
17 Jan 2025 - Inspecting ourselves to death -
Performance discovery
(2):
10 Jan 2025 - IOPS vs. IOPS
RECENT COMMENTS
-
By Dmitry on Answer: What does this code do?
-
By Pyth0n on Challenge: What does this code do?
-
By Oren Eini on Performance discovery: IOPS vs. IOPS
-
By kpvleeuwen on Performance discovery: IOPS vs. IOPS
-
By Oren Eini on Aggregating trees with RavenDB
Syndication
| Main feed | |
| Comments feed |
Comments
Just gave it a try - Voron.Tryout indeed works, but the unit tests end in a segfault or a deadlock - I'll take a look at this if I find some time over the weekend.
e-tobi, Yes, there seems to be something that we missed there.
Now you're talking. Been waiting for this port.
As an aside, have you had a chance to look at Aerospike? Worth a peek at what they're up to...
I'm feeling uncomfortably excited. I've wanted to use RavenDB on personal projects, but I use Linux almost exclusively.
@J Healy: Aerospike says it's an in-memory database that can handle 100TB of data on just a few servers. The biggest server I can get has 256GB of RAM. Just storing the raw data would take 400 of them. Indexing it would take more. They're almost certainly lying about it being an in-memory database. I'd guess they have some good caching algorithms and efficient point queries but are ultimately serving out of durable storage.
Clearly once they get over the memory capacity of a box they go hybrid...
"You can run Aerospike in pure RAM with spinning disks for persistence or as a hybrid memory database with RAM and flash.
Indexes (primary and secondary) are always stored in DRAM for fast access and are never stored on Solid State Drives (SSDs) to ensure low wear.
Unlike other databases that use the linux file system that was built for rotational drives, Aerospike has implemented a log structured file system to access flash – raw blocks on SSDs – directly. Access is optimized for how flash works – with small block reads and large block writes – and parallelized across multiple SSDs for better throughput.
Per namespace storage configuration – each namespace can be configured to store data on DRAM or on SSDs."
Comment preview