This topic introduces the terms that are related to Graph Compute.
instance
A Graph Compute instance that you create is an exclusive instance. Each instance has two replicas and is deployed in a cluster architecture that consists of proxies and searchers. You can create multiple graph schemas in an exclusive instance.
iGraph engine
The iGraph engine is independently developed by Alibaba Cloud. This engine provides high-performance and low-latency query and computing capabilities. The iGraph engine is the core of Graph Compute.
proxy and searcher
Proxies reside in the computing layer of the iGraph engine. Proxies receive user requests, convert the requests into execution plans, and then forward the execution plans to searchers. Proxies support complex operators such as merging and grouping.
Searchers reside in the storage layer of the iGraph engine. Searchers load and manage various indexes and process the query requests forwarded by proxies. Searchers support simple operators such as sorting, scattering, and truncation.
table
Graph Compute uses Key-Value (KV), PKey-SKey-Value (KKV), and inverted tables to support the data model of property graphs.
A table contains information such as source data, custom fields, and index-related schema information. The information allows Graph Compute to build indexes.
Graph Compute builds indexes by using an offline system and generates KV, KKV, and inverted index schemas for the corresponding table types. After the indexes are loaded into the iGraph engine, you can query data by calling API operations or using SDKs.
property graph
A property graph, as shown in Figure 1, has the following characteristics:
A property graph consists of vertices and edges.
User vertices: U1, U2, U3, and U4. Commodity vertices: I1, I2, and I3.
Friend relationship edges: friend1, friend2, friend3, and friend4. Purchase relationship edges: buy1, buy2, buy3, buy4, and buy5.
All edges in a property graph are directed edges.
All vertices and edges in a property graph have properties.
A user vertex has the following properties: name, age, and gender.
A commodity vertex has the following properties: title, price, and category.
A friend relationship edge has the following properties: time and score.
A purchase relationship edge has the buytime property.
Both vertices and edges have multiple types. Vertices or edges of the same type have the same properties.
KV table
A KV table, as shown in Figure 2, is often used to store the vertices of a property graph, such as the user and commodity vertices in Figure 1. The Key field stores the IDs of vertices, such as U1. The Value field stores the properties of vertices, such as the name, age, and gender properties of a user vertex.
KKV table
A KKV table, as shown in Figure 3, is often used to store the edges of a property graph, such as the friend relationship edges in Figure 1. The primary key (PKey) field stores the source vertex IDs of edges, such as U1. The secondary key (SKey) field stores the destination vertex IDs of edges, such as U2, U3, and U4. The Value field stores the properties of edges, such as the time and score properties of a friend relationship edge.
inverted table
An inverted table is also known as an index table. An inverted table stores the mappings between terms and document IDs. An Inverted table also defines the source data and data types, including the inverted configuration schema-related fields.
An inverted index is also known as a postings file or an inverted file. An inverted index is an indexing method that is used to store mappings from terms to the locations in a document or a set of documents. An inverted index allows fast full-text searches. An inverted index is the most commonly used data structure in document retrieval systems.
You can use inverted indexes to find the documents that contain a term and the locations of the term in the documents, and obtain other information for analysis, such as term frequency.
Gremlin syntax
iGraph implements open source Gremlin operators by using the C++ programming language and provides the query syntax.
MaxCompute
MaxCompute (formerly known as ODPS) is a fast and fully managed data warehousing solution that can process terabytes or petabytes of data. MaxCompute supports a comprehensive set of data import methods and a variety of classic distributed computing models.