Glossary

the term

Reading the conceptual explanation of related vocabulary below may help you understand our overall structure.

Terms	Definition
A
AST	AST is an abstract syntax tree, which is the tree structure representation of the code and is the basic part that composes the compiler's working mode
C
Cluster	The distributed deployment form of MatrixOne is composed of multiple hosts and logically forms a whole.
E
Explicit Transactions	Explicit Transactions, that is, a specified transaction, which transaction needs to be determined by yourself which batch of work must be completed successfully, otherwise all parts will not be completed. You can use the BEGIN TRANSACTION and ROLLBACK TRANSACTION or COMMIT TRANSACTION keywords for control.
I
Implicit transactions	Implicit transactions, that is, automatically commit transactions.
O
Optimistic transaction	Optimistic transaction, Optimistic transaction, that is, when the transaction starts, there will be no conflict detection or locking. The current relevant data will be cached to the corresponding memory area, and the data will be added, deleted and modified.
P
Pessimistic transaction	Pessimistic transaction, MatrixOne's default transaction mode, that is, when the transaction starts, it will assume that the transaction-related table is in a state where a write conflict will occur, lock the corresponding data table or data row in advance. After completing the locking action, the insertion, modification or deletion of the data is cached in memory. After committing or rolling back, the data will be dropped and the lock will be released.
S
Snapshot Isolation (SI)	Snapshot Isolation is a multi-version concurrent control technology widely used in practice. MatrixOne supports distributed transactions at the Snapshot isolation level.

Important Concept

Concept	Definition
A
Auto-Rebalance	In a distributed system, the automatic balancing process of storage volume and read and write load of multiple servers is called Auto-Rebalance.
C
Consistency	MatrixOne supports strong consistency, ensuring that after successfully writing data, the latest data can be read on any Store (node).
E
Execution Plan	The execution plan in the database is a graphical representation of the query operation generated by the query optimizer, and you can get the most efficient method to perform the operation
F
Fault-Tolerance	Fault-Tolerance means the ability of a system to continue running after one or more of its components fail.
M
Monolitic Engine	Monolitic Engine is a hyper-converged engine that can support hybrid workloads such as TP, AP, timing, and machine learning.
Materialized View	Materialized View is a materialized view, which is a pre-calculated data set, stored for subsequent use, and can usually improve the operation efficiency of the query.
Metadata	Metadata is metadata, which is data used to describe the structural information of the data in the database.
P
Paxos	Paxos is a consistency algorithm that maintains consistency between distributed computers that communicate asynchronously.
R
Raft	Raft is an easy-to-understand consistency protocol algorithm that is comparable to Paxos in terms of fault tolerance and performance.
Raft Group and Leader	Raft defines a leader and many followers in a group. A group represents a replication state machine, and only the leader can respond to client requests and will then be communicated to followers.
S
SIMD instruction	SIMD is the abbreviation of Single Instruction/Multiple Data, which is a single instruction and multiple data stream. SIMD operation generally refers to a calculation method that can process multiple data using one instruction.
T
Transaction	A series of operations performed in the database that meets the basic requirements of ACID.
TAE	Full name Transactional Analytic Engine, storage engine. The storage engine is the main public interface of the storage layer, which can support both row and column storage and transaction processing capabilities.
V
Vectorized Execution	Vectorized execution improves the speed of the analytical query engine by effectively utilizing the CPU's cache. Arrow's columnar format can use lightweight architectures such as dictionary encoding, bit packing and run length encoding, which further improves query efficiency.