joins Postgres Conference China 2016 中国用户大会 Global Transaction Manager (GTM) • Handles necessary MVCC tasks • Transaction IDs • Snapshots • Manages cluster wide values • Timestamps • Sequences • GTM Postgres Conference China 2016 中国用户大会 Multiversion Concurrency Control (MVCC) Postgres Conference China 2016 中国用户大会 MVCC  Readers do not block writers  Writers do not block readers  Transaction contain a list of running XIDs July 12th, 2012 Postgres-XC 50 Postgres Conference China 2016 中国用户大会 MVCC – Regular PostgreSQL Example: T1 Begin... INSERT… T2 Begin; INSERT...; Commit; T3

PostgreSQL maintains data consistency by using a multiversion model (Multiversion Concurrency Control, MVCC). This means that while querying a database each transaction sees a snapshot of data (a database version) isolation for each database session. The main advantage to using the MVCC model of concurrency control rather than locking is that in MVCC locks acquired for querying (reading) data do not conflict with locks facilities are also available in PostgreSQL for applications that cannot adapt easily to MVCC behavior. However, proper use of MVCC will generally provide better performance than locks. 12.2. Transaction Isolation

Internally, data consistency is maintained by using a multiversion model (Multiversion Concurrency Control, MVCC). This means that while querying a database each transaction sees a snapshot of data (a database version) transaction updates on the same data rows, providing transaction isolation for each database session. MVCC, by eschewing explicit locking methodologies of traditional database systems, minimizes lock contention performance in multiuser environments. The main advantage to using the MVCC model of concurrency control rather than locking is that in MVCC locks acquired for querying (reading) data do not conflict with locks

data consistency is maintained by using a multiversion model (Multiversion Concurrency Con- trol, MVCC). This means that while querying a database each transaction sees a snapshot of data (a database version) transaction updates on the same data rows, providing transaction isolation for each database session. MVCC, by eschewing explicit locking methodologies of traditional database systems, minimizes lock contention performance in multiuser environments. The main advantage to using the MVCC model of concurrency control rather than locking is that in MVCC locks acquired for querying (reading) data do not conflict with locks

PostgreSQL maintains data consistency by using a multiversion model (Multiversion Concurrency Control, MVCC). This means that while querying a database each transaction sees a snapshot of data (a database version) isolation for each database session. The main advantage to using the MVCC model of concurrency control rather than locking is that in MVCC locks acquired for querying (reading) data do not conflict with locks facilities are also available in PostgreSQL for applications that cannot adapt easily to MVCC behavior. However, proper use of MVCC will generally provide better performance than locks. 12.2. Transaction Isolation

Internally, data consistency is maintained by using a multiversion model (Multiversion Concurrency Control, MVCC). This means that while querying a database each transaction sees a snapshot of data (a database version) transaction updates on the same data rows, providing transaction isolation for each database session. MVCC, by eschewing explicit locking methodologies of traditional database systems, minimizes lock contention performance in multiuser environments. The main advantage to using the MVCC model of concurrency control rather than locking is that in MVCC locks acquired for querying (reading) data do not conflict with locks

Internally, data consistency is maintained by using a multiversion model (Multiversion Concurrency Control, MVCC). This means that while querying a database each transaction sees a snapshot of data (a database version) transaction updates on the same data rows, providing transaction isolation for each database session. MVCC, by eschewing explicit locking methodologies of traditional database systems, minimizes lock contention performance in multiuser environments. The main advantage to using the MVCC model of concurrency control rather than locking is that in MVCC locks acquired for querying (reading) data do not conflict with locks

data consistency is maintained by using a multiversion model (Multiversion Concurrency Con- trol, MVCC). This means that while querying a database each transaction sees a snapshot of data (a database version) transaction updates on the same data rows, providing transaction isolation for each database session. MVCC, by eschewing explicit locking methodologies of traditional database systems, minimizes lock contention performance in multiuser environments. The main advantage of using the MVCC model of concurrency control rather than locking is that in MVCC locks acquired for querying (reading) data do not conflict with locks

Internally, data consistency is maintained by using a multiversion model (Multiversion Concurrency Control, MVCC). This means that while querying a database each transaction sees a snapshot of data (a database version) transaction updates on the same data rows, providing transaction isolation for each database session. MVCC, by eschewing explicit locking methodologies of traditional database systems, minimizes lock contention performance in multiuser environments. The main advantage of using the MVCC model of concurrency control rather than locking is that in MVCC locks acquired for querying (reading) data do not conflict with locks

Internally, data consistency is maintained by using a multiversion model (Multiversion Concurrency Control, MVCC). This means that while querying a database each transaction sees a snapshot of data (a database version) transaction updates on the same data rows, providing transaction isolation for each database session. MVCC, by eschewing explicit locking methodologies of traditional database systems, minimizes lock contention performance in multiuser environments. The main advantage of using the MVCC model of concurrency control rather than locking is that in MVCC locks acquired for querying (reading) data do not conflict with locks