Glossary¶
Terms¶
These short definitions can help you build the right mental model when reading the rest of the docs.
Term |
Definition |
|---|---|
A |
|
AST |
Abstract Syntax Tree — the tree representation of code, a fundamental piece of a compiler. |
C |
|
Cluster |
The distributed deployment form of MatrixOne, made up of multiple hosts that act as a single logical system. |
E |
|
Explicit Transaction |
A transaction in which you explicitly decide which set of operations must succeed or fail together, controlled with |
I |
|
Implicit Transaction |
An auto-commit transaction. |
O |
|
Optimistic Transaction |
A transaction that does no conflict detection or locking when it starts; it caches the relevant data in memory and applies inserts / updates / deletes against it. |
P |
|
Pessimistic Transaction |
The default transaction model in MatrixOne. When the transaction starts, it assumes write contention is possible and locks the relevant tables or rows in advance; insert / update / delete are buffered in memory and persisted only on commit (or discarded on rollback), with the locks released. |
S |
|
Snapshot Isolation (SI) |
A widely used multi-version concurrency control technique. MatrixOne supports distributed transactions at Snapshot Isolation. |
Key concepts¶
Concept |
Definition |
|---|---|
A |
|
Auto-Rebalance |
In a distributed system, the automatic process of rebalancing storage and read/write load across multiple servers. |
C |
|
Consistency |
MatrixOne provides strong consistency, guaranteeing that once a write succeeds, the latest data is visible from every Store (node). |
E |
|
Execution Plan |
A graph representation of query operations produced by the query optimizer, describing the most efficient way to execute the query. |
F |
|
Fault-Tolerance |
The ability of the system to keep running after one or more components fail. |
M |
|
Monolithic Engine |
A hyper-converged engine that supports mixed workloads — TP, AP, time-series, machine learning. |
Materialized View |
A pre-computed dataset, persisted for later reuse, typically used to speed up queries. |
Metadata |
Data describing the structural information of data in a database. |
P |
|
Paxos |
A consensus algorithm that keeps a group of distributed computers communicating asynchronously in agreement. |
R |
|
Raft |
A consensus algorithm easier to understand than Paxos, comparable in fault tolerance and performance. |
Raft Group and Leader |
Raft defines a leader and several followers in a group. The group represents a replicated state machine; only the leader handles client requests, then propagates them to followers. |
S |
|
SIMD instruction |
Single Instruction / Multiple Data — a way to process multiple pieces of data with one instruction. |
T |
|
Transaction |
A series of operations against the database that satisfies the basic ACID requirements. |
TAE |
Transactional Analytic Engine — the storage engine. The storage engine is the main interface of the storage layer; TAE supports both row and columnar storage and transactional processing. |
V |
|
Vectorized Execution |
By making efficient use of CPU caches, vectorized execution speeds up analytical query engines. Arrow’s columnar format enables lightweight techniques such as dictionary encoding, bit packing, and run-length encoding, all of which further improve query efficiency. |