Meetup - YugabyteDB - Introduction and key features
0 likes100 views
YugabyteDB is a distributed SQL database that is PostgreSQL-compatible and designed for scalability and resilience, allowing for geo-distributed deployments. It utilizes automatic sharding and the Raft consensus algorithm to manage data consistency and fault tolerance across multiple nodes. Its architecture supports cloud-native applications and eliminates the need for learning a new database system, making it suitable for OLTP workloads and multi-cloud environments.
Meetup - YugabyteDB - Introduction and key features
- 1. © 2023 All Rights Reserved 1 YugabyteDB Introduction and key features Franck Pachot, Developer Advocate @FranckPachot
- 2. © 2023 All Rights Reserved Franck Pachot Developer Advocate at Yugabyte Past: 20+ years in databases, dev and ops, consulting Oracle ACE Director, AWS Data Hero Oracle Certified Master, AWS Database Specialty 2 [email protected] dev.to/FranckPachot @FranckPachot
- 3. © 2023 All Rights Reserved YugabyteDB 3 Distributed SQL database PostgreSQL compatible Open Source ⇨ Scale-Out (horizontal scalability) ⇨ Familiarity with the most popular DB for OLTP ⇨ Freedom, no lock-in, community Founders have worked on Cassandra and HBASE at Facebook Started based on Google Spanner paper
- 4. © 2023 All Rights Reserved NoSQL Monolithic RDBMS Future of "systems of record" databases (OLTP) Architectural resilience Horizontal scaling Geo-distribution ACID transactions SQL + RDBMS features SQL Distributed SQL
- 5. © 2023 All Rights Reserved On-Premises Datacenters Deploy Anywhere YugabyteDB architecture 5 Distributed, Transactional Storage Layer Automatic Sharding Load Balancing Distributed Transactions Raft Consensus Open Source Distributed SQL Database ⭐ https://github.com/yugabyte/yugabyte-db Raft consensus protocol geo distribution YSQL YCQL Pluggable Query Layer YCQL API Cassandra Compatible YSQL API Postgres Compatible Other APIs (future) PostgreSQL Application
- 6. © 2023 All Rights Reserved Sharding 6 Breaks up large tables and indexes to smaller pieces (around 1-10 GB) Each row in the user table (or index entry) gets mapped to exactly 1 shard Split by hash or range on the primary key (or index keys) Sharding in YugabyteDB is automatic, distributed across the nodes Sharding ------------
- 7. © 2023 All Rights Reserved Raft algorithm and LSM-Tree 7 Each tablet is a Raft group Each write go to the Leader, is replicated to Followers, wait for quorum Consistent read go to Leader (bounded stale reads from Followers) In case of failure, an new leader is elected from the quorum The tablet peer rows are stored in LSM-Tree (based on RocksDB)
- 8. © 2023 All Rights Reserved YugabyteDB deployment 8 On any Linux: VMs, Pods, containers Each node has a PostgreSQL (Backend, Planner, Executor) Each node stores some tablet peers Those are Tablet Servers with local storage and network access An additional "master" tablet holds the cluster metadata DocDB distributed storage & transactions YSQL stateless query layer Read/ Write Read/ Write Read/ Write Read/ Write Read/ Write Read/ Write
- 9. © 2023 All Rights Reserved 1. Single Region, Multi-Zone Availability Zone 1 Availability Zone 2 Availability Zone 3 Consistent Across Zones No WAN Latency But No Region-Level Failover/Repair 2. Single Cloud, Multi-Region Region 1 Region 2 Region 3 Consistent Across Regions with Auto Region-Level Failover/Repair 3. Multi-Cloud, Multi-Region Cloud 1 Cloud 2 Cloud 3 Consistent Across Clouds with Auto Cloud-Level Failover/Repair Deployment topology 9
- 10. © 2023 All Rights Reserved Quick Start (dev environment) 10 # Start YugabyteDB (3 nodes) docker network create -d bridge yb docker run -d --name yb0 --hostname yb0 --net=yb -p 5433:5433 -p7000:7000 yugabytedb/yugabyte:2.14.6.1-b4 yugabyted start --daemon=false --listen yb0 docker run -d --name yb1 --hostname yb1 --net=yb -p 5434:5433 yugabytedb/yugabyte:2.14.6.1-b4 yugabyted start --daemon=false --listen yb1 --join yb0 docker run -d --name yb2 --hostname yb2 --net=yb -p 5435:5433 yugabytedb/yugabyte:2.14.6.1-b4 yugabyted start --daemon=false --listen yb2 --join yb0 # Connect to any nodes psql postgres://localhost:5433 -- Web console on http://localhost:7000
- 11. © 2023 All Rights Reserved 11 Q & A ● Slack: https://communityinviter.com/apps/yugabyte-db/register ● Github: https://github.com/yugabyte ● Newsletter: https://www.yugabyte.com/newsletter
- 12. © 2023 All Rights Reserved 12 Reason for Distributed SQL - scale-out for elasticity and resilience - rolling upgrades, geo-distribution Reason for PostgreSQL compatibility - no need to learn a new DB Reason for a fork of PostgreSQL - the best compatibility with advanced features Core messages
- 13. © 2023 All Rights Reserved 13 Major difference with PostgreSQL: - think about the primary key (hash/range sharding) - leverage new access patterns (primary index, skip scan, index only scan on secondary index ) - don't worry about vacuum Core messages
- 14. © 2023 All Rights Reserved 14 When it can be an alternative to PostgreSQL: - OLTP (with some analytics push-downs) - cloud native (Compute instances, kubernetes pods,... all active) - multitenant (no hardware limitation to scale) - geo-distribution (data residency) Core messages
- 15. © 2023 All Rights Reserved 15 E-mail: [email protected] Blogs: dev.to/FranckPachot blog.yugabyte.com/author/fpachot Twitter: @FranckPachot Youtube: youtube.pachot.net Twitch: www.twitch.tv/franckpachot LinkedIn: www.linkedin.com/in/franckpachot Q & A ● Slack: https://communityinviter.com/apps/yugabyte-db/register ● Github: https://github.com/yugabyte ● Newsletter: https://www.yugabyte.com/newsletter