性能优化 #

优化目标：
├── 提高查询响应速度
├── 提高吞吐量
├── 降低资源消耗
├── 提高并发能力
└── 优化用户体验

PROFILE MATCH (p:Person)-[:KNOWS]->(f:Person)
WHERE p.name = 'Tom'
RETURN p, f

EXPLAIN MATCH (p:Person)-[:KNOWS]->(f:Person)
WHERE p.name = 'Tom'
RETURN p, f

CREATE INDEX FOR (p:Person) ON (p.name)

MATCH (p:Person)
WHERE p.name = 'Tom'
RETURN p

MATCH (p:Person)
RETURN p
LIMIT 100

:param name => 'Tom'
MATCH (p:Person)
WHERE p.name = $name
RETURN p

MATCH (p:Person {name: 'Tom'})-[:KNOWS]->(f:Person)
RETURN f

MATCH (p:Person)
WHERE p.status = 'active'
WITH p LIMIT 1000
MATCH (p)-[:KNOWS]->(f:Person)
RETURN p, f

MATCH (p:Person)
USING INDEX p:Person(name)
WHERE p.name = 'Tom'
RETURN p

CREATE INDEX FOR (p:Person) ON (p.email)
CREATE INDEX FOR (p:Person) ON (p.name, p.age)

CREATE INDEX FOR (p:Person) ON (p.lastName, p.firstName)

MATCH (p:Person)
WHERE p.lastName = 'Smith' AND p.firstName = 'John'
RETURN p

CREATE FULLTEXT INDEX person_search FOR (p:Person) ON EACH [p.name, p.description]

CALL db.index.fulltext.queryNodes('person_search', 'engineer')
YIELD node, score
RETURN node.name, score

SHOW INDEXES YIELD name, state, populationPercent

建议：
├── 只为常用查询属性创建索引
├── 每个标签不超过5个索引
├── 定期审查索引使用情况
└── 删除未使用的索引

CREATE (p:Person:Employee {name: 'Tom'})

CREATE (a)-[:KNOWS {since: 2020, level: 'close'}]->(b)

MATCH (a)-[:KNOWS*1..3]->(b)
RETURN a, b

CREATE (p:Person {
    name: 'Tom',
    age: 30,
    createdAt: datetime(),
    tags: ['developer', 'designer']
})

超级节点问题：
├── 一个节点有大量关系
├── 查询性能下降
├── 内存消耗大

解决方案：
├── 分拆节点
├── 使用中间节点
├── 限制关系数量
└── 使用缓存

server.memory.heap.initial_size=4G
server.memory.heap.max_size=4G

server.memory.pagecache.size=2G

CALL dbms.queryJmx('org.neo4j:*') YIELD name, attributes
WHERE name CONTAINS 'Memory'
RETURN name, attributes

dbms.transaction.timeout=30s

UNWIND range(1, 10000) AS id
CALL {
    WITH id
    CREATE (p:Person {id: id})
} IN TRANSACTIONS OF 1000 ROWS

db.tx_log.rotation.retention_policy=2 days 1G size

server.bolt.thread_pool_max_size=200
server.bolt.thread_pool_min_size=50

CALL dbms.setConfigValue('dbms.logs.query.threshold', '1s')

MATCH (p:Person)
WHERE p.status = 'active'
WITH p
CALL {
    WITH p
    MATCH (p)-[:KNOWS]->(f:Person)
    RETURN f
} RETURN p, f

server.logs.query.enabled=true
server.logs.query.threshold=1s

CALL dbms.listQueries()

CALL dbms.killQuery('query-id')

CALL dbms.queryJmx('org.neo4j:*')

server.metrics.enabled=true
server.metrics.prometheus.enabled=true
server.metrics.prometheus.endpoint=0.0.0.0:2004

dbms.jvm.additional=-Dcom.sun.management.jmxremote
dbms.jvm.additional=-Dcom.sun.management.jmxremote.port=3637
dbms.jvm.additional=-Dcom.sun.management.jmxremote.authenticate=false
dbms.jvm.additional=-Dcom.sun.management.jmxremote.ssl=false

建议：
├── 使用PROFILE分析查询
├── 使用索引加速查询
├── 限制返回数据量
├── 使用参数化查询
├── 避免全图扫描
├── 合理使用WITH
└── 使用批量操作

建议：
├── 合理设计标签和关系
├── 避免过深的遍历
├── 处理超级节点
├── 使用适当的属性类型
└── 定期清理无用数据

建议：
├── 合理配置内存
├── 设置事务超时
├── 启用慢查询日志
├── 监控性能指标
└── 定期维护数据库

MATCH (p:Person)-[:KNOWS*1..5]->(f:Person)
WHERE p.name = 'Tom'
RETURN f

MATCH (p:Person {name: 'Tom'})-[:KNOWS*1..3]->(f:Person)
RETURN f
LIMIT 100

UNWIND range(1, 100000) AS id
CREATE (p:Person {id: id})

UNWIND range(1, 100000) AS id
CALL {
    WITH id
    CREATE (p:Person {id: id})
} IN TRANSACTIONS OF 1000 ROWS

MATCH (p:Person)-[:KNOWS]->(f:Person)
RETURN p.name, count(f)

MATCH (p:Person)
WHERE exists((p)-[:KNOWS]->())
WITH p
MATCH (p)-[:KNOWS]->(f:Person)
RETURN p.name, count(f)