YDB
YDB is an open-source Distributed SQL Database that combines high availability and scalability with strong consistency and ACID transactions.
True Elastic Scalability
Add or remove nodes to easily scale up and scale down as needed. YDB is proven to work in real production with millions of transactions per second and petabytes of data for mission-critical, real-time applications.
Fault-tolerant
YDB is designed to work in three availability zones, ensuring availability even if a node or availability zone goes offline.
Easy to use
YDB combines strong consistency, ACID transactions, high performance queries, fast data ingest with a familiar SQL dialect and JSON API support. Works with any modern workloads: key-value, relational, JSON.
Automatic disaster recovery
Automatic recovery after a disk, server, or even a data center fails, with minimum latency disruptions for applications.
Available in any cloud
YDB is available for self-deployment, including Kubernetes, on-premise and cloud environments.
Open Source
Licensed under Apache 2.0.
No risk of cloud or vendor lock-in.
Success stories
Market Cart
The Cart is one of the key components of any marketplace or online store.
Using YDB as a database allowed Market Cart to withstand a hundredfold increase in the load on the Cart, while observing strict guarantees for response times.
Moreover, the migration was completed by just one developer in one month.
How it works
YDB architecture
We use commodity hardware and shared-nothing architecture, disaggregated compute and storage layers, and build a system based on logical components — tablets.
Hierarchy
Similar to a file system tables could be organized into a hierarchy using directories.
Table
YDB provides users with a familiar abstraction: tables. Tables must contain a primary key, the data is sorted by the primary key. Tables are automatically sharded by primary key range by size or load.
Split by load
The tablet will automatically split when the load increases.
.png)
Split by size
The tablet will automatically split when the size increases.
Automatic balancing
YDB evenly distributes tablets among the nodes, and moves loaded tablets from loaded nodes. CPU, Memory, Network metrics are tracked.
Distributed Storage Internals
We write all the code for working with block devices ourselves. The PDisk component is responsible for working with the block device. Above PDisk is the VDisk abstraction layer. There is a special component — DSProxy between the tablet (part of the table) and VDisk. DSProxy analyzes disk availability and characteristics and depending on it can make a decision to exclude the disk from work.
DSProxy
YDB writes data to 3 Availability Zones, doesn’t send requests to obviously bad disks, and continues to operate without interruption even if one AZ and a disk in another AZ are lost.
How to start
Docker
Pull the current public version of the Docker image:
docker pull cr.yandex/yc/yandex-docker-local-ydb:latest
Create a working directory and start the container from it:
docker run -d --rm --name ydb-local -h localhost -p 2135:2135 -p 8765:8765 -p 2136:2136 -v $(pwd)/ydb_certs:/ydb_certs -v $(pwd)/ydb_data:/ydb_data -e YDB_DEFAULT_LOG_LEVEL=NOTICE -e GRPC_TLS_PORT=2135 -e GRPC_PORT=2136 -e MON_PORT=8765 cr.yandex/yc/yandex-docker-local-ydb:latest
Go to the Getting started — Self-hosted deploy — Docker in the YDB documentation to get detailed information.
