Slide 1
Slide 1 text
Sharding in modern
[No]SQL databases
Konstantin Osipov, CTO, Tarantool
Novosibirsk, Russia, 01/04/2018
Slide 2
Slide 2 text
whoami
● MySQL core team member 2003-2010
● Tarantool core team member 2010-now
● http://t.me/tarantoolru
● http://tarantool.org
Slide 3
Slide 3 text
The plan
NoSQL системы всё больше смещаются в сторону NewSQL.
Давайте посмотрим, как на это влияют ключевые технические решения в
реализации шардинга, принятые на старте проекта.
● sharding introduction
● vBucket or not; hash or range!
● routing
● secondary keys are hard
Slide 4
Slide 4 text
Sharding: introduction
Sharding - горизонтальное разбиение данных СУБД по нескольким
машинам. Позволяет линейно масштабировать ёмкость и (возможно)
пропускную способность СУБД. Требует решения 3 задач:
● выбрать принцип разбиения данных
● обеспечить хранение метаданных
● организовать роутинг запросов к данным
Slide 5
Slide 5 text
Sharding: shard function
Три основных способа задания shard function:
● Hash
● Range
● Range* (Stonebraker)
Shard-функция также влияет на алгоритм балансировки: может
перемещаться ключ, диапазон (range) или использоваться концепция
virtual buckets.
Slide 6
Slide 6 text
Sharding: consistent hash
Slide 7
Slide 7 text
Sharding: Guava
Slide 8
Slide 8 text
Sharding: hash + virtual buckets in Couchbase
Slide 9
Slide 9 text
Sharding: ranges and chunks in MongoDB
Slide 10
Slide 10 text
Sharding: replica sets in MongoDB
Slide 11
Slide 11 text
Sharding: chunk splits and migrations in MongoDB
Slide 12
Slide 12 text
Sharding: ranges in CockroachDB
Slide 13
Slide 13 text
mongodb
For queries that don’t include the shard key, mongos must query all shards, wait
for their response and then return the result to the application. These
“scatter/gather” queries can be long running operations.
However, range based partitioning can result in an uneven distribution of
data, which may negate some of the benefits of sharding. For example, if the
shard key is a linearly increasing field, such as time, then all requests for a
given time range will map to the same chunk, and thus the same shard. In this
situation, a small set of shards may receive the majority of requests and the
system would not scale very well.
Slide 14
Slide 14 text
Sharding: range* schema in VoltDB & Tarantool
Slide 15
Slide 15 text
Sharding: metadata & routing
Метаданные шардинга: определения таблиц и таблица роутинга
● На выделенных узлах - e.g. mongodb config server
● Metadata is data - Couchbase, CockroachDB, Tarantool
От решения задачи хранения метаданными зависит архитектура
роутинга:
● на клиенте
● на выделенных узлах
● на storage узлах
Slide 16
Slide 16 text
Sharding: MongoDB example
Slide 17
Slide 17 text
Sharding: Couchbase example
Slide 18
Slide 18 text
Sharding: cluster topology in Couchbase
Slide 19
Slide 19 text
Summary: design choices
Couchbase MongoDB CockroachDB VoltDB | Tarantool
Shard function Hash + vbuckets Hash, range in
2.x, vbucket
Range Range* + vbucket
Metadata Metadata is data Config server Metadata is data* Metadata is data
Routing Client and server Router Server Server
Non-sharded
tables
No Yes No Yes
Secondary keys Global and local Local Global only Local, Materialized
views*
Slide 20
Slide 20 text
Summary: hash vs. range based sharding
Hash Range
Write запросы/time series Linear scaling Hotspots
Primary key read Linear scaling Linear scaling
Partial key read Map/reduce Linear scaling
Range read Map/reduce Linear scaling
Chunk splits and migration Easy Difficult
Metadata size Small Large
Slide 21
Slide 21 text
Summary: which database to choose?
Slide 22
Slide 22 text
Links
https://dzone.com/articles/principles-of-sharding-for-relational-databases
http://www.couchbase.com/binaries/content/assets/website/docs/whitepapers/couchbase-server-under-the-hood.pdf
http://www.oracle.com/technetwork/database/availability/oraclesharding-whitepaper-3675509.pdf
https://github.com/VoltDB/voltdb/blob/master/src/frontend/org/voltdb/Buckets.java#L30
https://docs.voltdb.com/tutorial/Part3.php
https://docs.voltdb.com/UsingVoltDB/DesignPartition.php
https://www.cockroachlabs.com/blog/automated-rebalance-and-repair/
https://www.cockroachlabs.com/blog/sql-in-cockroachdb-mapping-table-data-to-key-value-storage/
https://developer.couchbase.com/documentation/server/3.x/developer/dev-guide-3.0/buckets.html
https://developer.couchbase.com/documentation/server/3.x/admin/Concepts/concept-vBucket.html
https://developer.couchbase.com/documentation/server/4.1/concepts/buckets-vbuckets.html
https://developer.couchbase.com/documentation/server/3.x/admin/Tasks/tasks-rebalance.html
https://docs.mongodb.com/manual/sharding