Transparent sharding with Spider: what's new and getting started
3 likes1,092 views
OpenWorks 2019 Session MariaDB Server 10.3 introduced transparent, built-in sharding with the Spider storage engine to scale out reads, writes and storage. MariaDB Server 10.4 will include a number of improvements, including DDL pushdown. In this session, Ralf Gebhardt and Kentoku Shiba of MariaDB show how to set up a sharded MariaDB cluster and scale out on demand, as well explore as best practices for high availability and consistency in a sharded deployment.
Ad
More Related Content
What's hot (20)
Similar to Transparent sharding with Spider: what's new and getting started (20)
Ad
More from MariaDB plc (20)
Ad
Recently uploaded (20)
![Download Wondershare Filmora Crack [2025] With Latest](https://cdn.slidesharecdn.com/ss_thumbnails/neo4j-howkgsareshapingthefutureofgenerativeaiatawssummitlondonapril2024-240426125209-2d9db05d-250419-250428115407-a04afffa-thumbnail.jpg?width=560&fit=bounds)
Transparent sharding with Spider: what's new and getting started
- 1. Transparent Sharding with Spider Kentoku Shiba Principal Software Engineer MariaDB Corporation Ralf Gebhardt Product Manager MariaDB Corporation
- 2. Agenda ● MariaDB Platform X3 ● Scalability, the big picture including Spider ● Spider - The Storage Engine ○ Sharding ○ High Availability ○ Use Case ○ Getting started ○ What’s new
- 4. Applications Containers MariaDB Platform X3 MariaDB MaxScale 2.3 CDC MariaDB Server 10.3 InnoDB/Spider/... MariaDB Server 10.3 ColumnStore 1.3 Transactional Analytical Kubernetes (Helm) Docker (Compose) C JDBC ODBC Node.js Ingest streaming data Kafka connector Administration SQL Diagnostic Manager SQLyog MariaDB Backup MariaDB Flashback Import bulk data Spark connector C/Java/Python API
- 5. Scalability The big picture and Spider
- 6. Application Connection 1 MariaDB Server Row storage MariaDB Server Columnar storage MariaDB MaxScale Transactional Analytical
- 7. MariaDB Server (Spider) Application Connection 1 MariaDB Server (ColumnStore) Node 1 Columnar storage Node 2 Columnar storage Node n Columnar storage Distributed storage MariaDB MaxScale Transactional Analytical MariaDB Server Row storage
- 8. MariaDB Server (Spider) Application Connection 1 MariaDB Server 1 Row storage MariaDB Server 2 Row storage MariaDB Server n Row storage MariaDB Server (ColumnStore) Sharding MariaDB MaxScale Transactional Analytical MariaDB Server Columnar storage
- 9. MariaDB Server (Spider) Application Connection 1 MariaDB Server 1 Row storage MariaDB Server 2 Row storage MariaDB Server n Row storage MariaDB Server (ColumnStore) CS node 1 Columnar storage CS node 2 Columnar storage CS node n Columnar storage Sharding Distributed storage MariaDB MaxScale Transactional Analytical
- 11. When and Why Should I Use Spider? ● A need to scale for OLTP ○ Data set is growing ○ Lots of concurrent operations on the data ● Sharding with Spider is simple ○ Transparent to the Application ○ Same optimisation rules apply as for using SQL partitioning ● No expensive hardware needed
- 12. What is Spider? ● A storage engine plugin ○ Spider doesn’t itself store data ● Manage storage and retrieval of data stored using other storage engines ● Sharding solution that stores data remotely on other servers ○ Distribute I/O-Load ● Partition tables using the Partition Engine ● View the data as if it is local MariaDB Server MariaDB Server 1 Sharding Application Engine=InnoDB MariaDB Server 2 Engine=InnoDB MariaDB Server 3 Engine=InnoDB Engine=Spider
- 13. What is Spider? 1. Receive a request 2. Execute the request a. Distribute SQL to data nodes b. Receive and consolidate results from data nodes 3. Send reply MariaDB Server MariaDB Server 1 Sharding Application Engine=InnoDB MariaDB Server 2 Engine=InnoDB MariaDB Server 3 Engine=InnoDB Engine=Spider2a 2b 31
- 15. Sharding using Partitions ● Usual rules to partitioning apply ● Goal - Reduce the data set for queries ● Partitioning rules ○ Range ○ Key ○ Hash ○ List Partition Engine Data Node Spider Data Node T1 P1 P2 P3 T1 T1 Data Node T1 XA 2 PC P1 P2 P3
- 16. Sharding using Partitions ● Engine condition pushdown ● Index hints ● Join ● Aggregation ● Direct update/delete Spider pushdown Data Node Spider Data Node T1 P1 P2 P3 T1 T1 Data Node T1 XA 2 PC P1 P2 P3
- 18. ● Full copy of the table on each data node ● For SELECTs, Spider performs load balancing ● INSERTs, UPDATEs and DELETEs are parallelized High Availability Data Node Spider T1 T1 Data Node T1
- 19. ● Full copy of a partition on a second data node ● For SELECTs, Spider performs load balancing ● INSERTs, UPDATEs and DELETEs are parallelized High Availability for Shards Spider T1 P1 P2 Data Node T1 P1 Data Node T1 P2 Data Node T1 P1 Data Node T1 P2
- 20. ● Multiple Spider Nodes can access the same Data Nodes ● Access Spider Node via ○ MariaDB MaxScale ○ MariaDB Connector/J Multi-Connection ○ Virtual IP setup High Availability for Spider Spider T1 P1 P2 Data Node T1 P2 Data Node T1 P1 Data Node T1 P2 Spider T1 P1 P2 Data Node P1 T1 P1 MaxScale / Connector/J / VIP
- 21. Use Case
- 22. Backup and Reporting Data Node Spider Data Node T1 P1 P2 P3 T1 T1 Data Node T1 P1 P2 P3 Slave T1 P1 P2 P3 Statement based Replication Application Backup / Report
- 23. Local Access to Data Nodes Prod1 T1 Machine 1 Production Monitoring Spider P 1 P 2 C Machine 2 Prod2 T2 ● Data Nodes for storing data of interest ○ Manufacturing ○ Per Location ○ Per Supplier ● Spider Node to ○ Monitor overall production status ○ Consolidated view of all locations ○ Supplier Chain, Dependencies, ...
- 25. Prepare Data Nodes $ mysql –u root … MariaDB> CREATE DATABASE dbsharded; MariaDB> CREATE USER 'spider'@'192.168.56.20' IDENTIFIED BY 'spider'; MariaDB> GRANT ALL ON dbsharded.* TO 'spider'@' 192.168.56.20'; ● Create database ● Create User for Spider Node ● Grant access to the shared database Maria DB Serve r
- 26. Prepare Spider Node $ mysql -u root -p < /usr/share/mysql/install_spider.sql ● Install the Spider storage engine $ mysql -u root -p MariaDB> SELECT engine, support, transactions, xa FROM information_schema.engines; +--------------------+---------+--------------+------+ | engine | support | transactions | xa | +--------------------+---------+--------------+------+ | SPIDER | YES | YES | YES | | CSV | YES | NO | NO | Maria DB Serve r
- 27. Prepare Spider Node ● Create the server definitions for the data nodes $ mysql -u root -p MariaDB> CREATE SERVER data1 FOREIGN DATA WRAPPER mysql OPTIONS (HOST '192.168.56.21', DATABASE 'dbsharded', USER 'spider', PASSWORD 'pw_spider'); MariaDB> CREATE SERVER data2 FOREIGN DATA WRAPPER mysql OPTIONS (HOST '192.168.56.22', DATABASE 'dbsharded', USER 'spider', PASSWORD 'pw_spider'); MariaDB> CREATE SERVER data3 FOREIGN DATA WRAPPER mysql OPTIONS (HOST '192.168.56.23', DATABASE 'dbsharded', USER 'spider', PASSWORD 'pw_spider'); Maria DB Serve r
- 28. Create Tables on Data Node ● Create tables on all data nodes or use existing ones ○ These tables will hold the data ● The Spider node needs to have access to the tables MariaDB> CREATE TABLE dbsharded.mytable( id INT NOT NULL, code VARCHAR(10), PRIMARY KEY(id) )ENGINE=INNODB; Maria DB Serve r
- 29. Create Table on Spider Node ● Create the table with the same structure on the Spider node ○ The table will not hold any data MariaDB> CREATE TABLE mytable(id INT NOT NULL, code VARCHAR(10), PRIMARY KEY(id)) ENGINE=SPIDER COMMENT='user "backend", table "mytable"' PARTITION BY RANGE(id) ( PARTITION p1 VALUES LESS THAN (100000) COMMENT='server "data1"', PARTITION p2 VALUES LESS THAN (200000) COMMENT='server "data2"', PARTITION p3 VALUES LESS THAN MAXVALUE COMMENT='server "data3"'); Maria DB Serve r Maria DB Serve r
- 30. Spider Management ● When installing Spider, new tables will be added to the system tables MariaDB> show tables like 'spider%'; +---------------------------+ | Tables_in_mysql (spider%) | +---------------------------+ | spider_link_failed_log | | spider_link_mon_servers | | spider_tables | | spider_xa | | spider_xa_failed_log | | spider_xa_member | +---------------------------+ Maria DB Serve r Maria DB Serve r
- 31. Spider Management ● When installing Spider, it will add ○ new tables to the system tables ○ New system variables ○ New status values ○ Spider UDFs MariaDB> show tables like 'spider%'; ... MariaDB> show global variables like 'spider%’; ... MariaDB> show global status like 'spider%’; Maria DB Serve r Maria DB Serve r
- 32. Good to know ● Check your usual queries for how they use partitions ● Define the partition of interest MariaDB> EXPLAIN PARTITIONS SELECT * FROM orders WHERE id < 150000; +------+-------------+--------+------------+ -------+---------------+---------+---------+------+------+-------------+ | id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | Extra | +------+-------------+--------+------------+ -------+---------------+---------+---------+------+------+-------------+ | 1 | SIMPLE | orders | p2 | range | PRIMARY | PRIMARY | 4 | NULL | 2 | Using where | +------+-------------+--------+------------+ -------+---------------+---------+---------+------+------+-------------+ MariaDB> SELECT * FROM orders PARTITION (p3) WHERE code = "JFK"; Maria DB Serve r Maria DB Serve r
- 34. What’s New in Spider? ● DDL Pushdown ○ Create table pushdown ● Support more aggregate functions for direct join feature ○ count(distinct) ○ sum(distinct) ○ avg() and avg(distinct) ● Support left and right join for direct join feature ● Options to replication ○ Slave trx isolation ● Options to quick mode ○ Use disk by size of result set from data nodes ● Contributions from Tencent
- 35. What’s New in Spider? DDL Pushdown ● Pushdown CREATE TABLE statement to data nodes. Data Node Spider Node Data Node Create table t1( Id int not null, … ) engine=InnoDB; Create table t1( Id int not null, … ) engine=InnoDB; Create table t1( Id int not null, … ) engine=Spider Partition by … ( Partition … Partition … ); 1 2 2 3
- 36. What’s New in Spider? Support more aggregate functions for direct join feature ● count(distinct) ● sum(distinct) ● avg() and avg(distinct) Data Node Spider Node Data Node Select count(distinct c2) from t1 where c1 = 1; 1 2Select count(distinct c2) from t1 where c1 = 1;
- 37. What’s New in Spider? Support left and right join for direct join feature Data Node Spider Node Data Node 1 2 Select … from t1 left join t2 on … where t1.c2 = t2.c1; Select … from t1 left join t2 on … where t1.c2 = t2.c1;
- 38. What’s New in Spider? Options to replication ● Slave trx isolation Data Node Slave (Spider Node) Data Node replication Master Repeatable read Repeatable read normally Read committed only for replication (spider_slave_trx_isolation = 1)
- 39. What’s New in Spider? Options to quick mode ● Use disk by size of result set from data nodes Data NodeData Node Spider Node Use a disk temporary table if a resultset from a data node is too big.
- 40. What’s New in Spider? Contributions from Tencent: ● Bug fixes
- 42. What’s Ahead for Spider? ● Vertical Partition (VP) Engine ○ Multi-dimensional sharding ○ VP merges multiple child tables into a single view ○ VP efficiently chooses child tables for each query ● DDL pushdown ○ Support synchronization between Spider nodes ○ Support other DDLs like alter table
- 43. THANK YOU!