Smartface, technologies and decay
@antonrevyako
Post #1122
1.87K
There will be no singularity Сейчас все пилят тиктоки свои БД. Помните про firebolt , да? Подписчик поделился ссылкой на проект своих друзей - unum.am Говорят, уделывает FoundationDB и все остальные noSQL. Правда, cloud only, без опенсорсов. Еще наткнулся на новую графовую базу - memgraph.com…
Remember these guys?
The guys from Unum have spent 2022 polishing their Key-Value Store, integrating SPDK, and starting an Open-Source ecosystem around it! The project UKV brings a shared compatibility layer between LevelDB, RocksDB, and their persistent and in-memory Key-Value Stores, some of which are also public.
- UKV can wrap any one of those Stores into a Standalone server with an Apache Arrow Flight RPC protocol, making it available across the network.
- Similar to Redis, with RedisGraph, RedisJSON, and RediSearch, UKV add supports for Graphs, Documents, and Vectors/Features. Unlike Redis, you still get ACID transactions and Strictly Serializable Consistency Guarantees. Document collections, for example, support JSON, BSON, MessagePack, and contents, as well as JSON-Pointer, JSON-Patch, and JSON-MergePatch RFC standards. Also, with the Embedded variant, you don't have to pay for Networking.
- UKV also brings structured bindings for high-level bindings. Python Graph bindings, for example, look like NetworkX. Python Tabular interface aims to become compatible with Pandas over time. In both cases, the query will operate on the entire dataset stored on disks, which can reach 100 TB+.
In 2023 guys are promising to work on replication and sharding, so stay tuned!
Designing the fastest ACID Key-Value Store / Ashot Vardanian (Unum cloud): https://www.youtube.com/watch?v=ybWeUf_hC7o
PS: The author is among us, you can ask questions in the comments to this post
The guys from Unum have spent 2022 polishing their Key-Value Store, integrating SPDK, and starting an Open-Source ecosystem around it! The project UKV brings a shared compatibility layer between LevelDB, RocksDB, and their persistent and in-memory Key-Value Stores, some of which are also public.
- UKV can wrap any one of those Stores into a Standalone server with an Apache Arrow Flight RPC protocol, making it available across the network.
- Similar to Redis, with RedisGraph, RedisJSON, and RediSearch, UKV add supports for Graphs, Documents, and Vectors/Features. Unlike Redis, you still get ACID transactions and Strictly Serializable Consistency Guarantees. Document collections, for example, support JSON, BSON, MessagePack, and contents, as well as JSON-Pointer, JSON-Patch, and JSON-MergePatch RFC standards. Also, with the Embedded variant, you don't have to pay for Networking.
- UKV also brings structured bindings for high-level bindings. Python Graph bindings, for example, look like NetworkX. Python Tabular interface aims to become compatible with Pandas over time. In both cases, the query will operate on the entire dataset stored on disks, which can reach 100 TB+.
In 2023 guys are promising to work on replication and sharding, so stay tuned!
Designing the fastest ACID Key-Value Store / Ashot Vardanian (Unum cloud): https://www.youtube.com/watch?v=ybWeUf_hC7o
PS: The author is among us, you can ask questions in the comments to this post




