时间:2021-07-01 10:21:17 帮助过:12人阅读
TigerGraph图
https://github.com/geektcp/tigergraph-gql
图数据库目前发展到第三代
第一代以neo4j为代表
第二代以Amazon Neptune为代表
第三代以tigergraph,arangodb为代表
gadmin: 管理命令,类似mysqladmin
gbar: 备份和存储命令,backup and resotre
GSQL Shell: 用于执行sql语句,类似mysql数据库的mysql命令
GraphStudio UI: 可视化控制台
GPE:图数据库计算引擎
Graph Store: 内置的一个基于内存的数据存储组件
DICT : 数据分片,类似arangodb的内部的分片机制
GSE : 存储引擎,GSE通过GPE接收操作数据,对Graph Store进行增删改查
GSQL Language:图数据库SQL,类似arangodb的aql
GSQL: 一个内部接口,GSQL Shell调用了这个接口执行图SQL
HA: tigergraph支持高可用
IDS: GSE的内部组件,用于把顶点表、边表转换成图
IUM: 安装,升级,维护
MultiGraph: 复杂图架构,支持多个子图
Native Parallel Graph:并行图架构,支持并行存储和分析,高并发,高伸缩性
Nginx :tigergraph的前端
REST++ or RESTPP:rest api,增加、删除和查询,没有编辑
SSO :
TigerGraph Platform
TigerGraph System
Kakfa
Zookeeper
tiger的边分为有向边,无相边。
有向边和无相边的区别是:界面上有向边有明显箭头,无向边就是直线。
有向边又有单向边,双向边。
无向边由于指定了边的起点类型和终点类型,而任何顶点的类型只能有一个,无所无向边其实还是有方向的。
只有起点和终点类型都是*的时候,才是真正的无向边。
单向边无法反向遍历。
无相边和双向边都可以反向遍历。
无向边的反向遍历结果相同。
双向边的反向遍历结果中边的类型不同。
双向边其实无向边效果差不多。
双向边和有向边的本质差别在于查询方向。双向边可以指定两个方向查询,有向边根本不能反方向查询。
原因是tiger的查询语法范式定死了格式。
SELECT t FROM vSetVarName:s – ((eType1|eType2):e) -> (vType1|vType2):t
第一个横杠-前面必须是顶点集合(冒号后面是别名,别名可以省略)
依据这个范式,最精简的写法是:
source = SELECT s FROM source:s -()-> ;
举例说明:
source = SELECT t FROM source:s -(:e)-> :t
上面的写法无法写成(因为dest是顶点集合,不是顶点类型):
dest = SELECT t FROM :s -(:e)-> dest:t
arangodb里面的边都是有向边,但是遍历方向是可以指定的,也是支持反向遍历。
FOR vertext[] IN []
FOR v,e,p in[1..3]
OUTBOUND|INBOUND|ANY startVertex [EDGES]
FILTER ...
tiger的GSQL里面要实现arangodb的AQL中的设定方向的效果,最好的方式是创建边的时候就定下来这条边是单向边还双向边。如果是单向边,创建边的时候就定好了的出的还是进的,即OUTBOUND还是INBOUND的。
如果创建边的时候指定双向边,那么遍历就是ANY方式。
在tiger的query里面不能再直接指定边的方向性了。
另外,如果都是双向边,这是遍历默认相当于arangodb的ANY方向,如何指定方向呢?
groupadd graph
useradd graph -g graph
visudo(添加如下一行)
graph ALL=(ALL) NOPASSWD:ALL
su - graph
wget http://dl.tigergraph.com/developer-edition/tigergraph-2.2.4-developer-patch.tar.gz
tar -zxvf tigergraph-2.2.4-developer-patch.tar.gz
cd tigergraph-2.2.4-developer
sudo ./install.sh (填入系统账号,比如当前使用graph用户,再填入安装路径即可)
官方没有提供卸载方式
建议先停掉tiger:
gadmin stop
然后注释掉crontab任务。
然后root用户下重新执行:
./install.sh -u tiger -p tiger -r /home/tiger/
./install.sh -l 40dd21b0acef8f25eca40f6a3077d458db09ea5925ba7da56894d4dcd8928d9c9340876800
gadmin --set license.key 40dd21b0acef8f25eca40f6a3077d458db09ea5925ba7da56894d4dcd8928d9c9340876800
关闭tigergraph:
gadmin stop
启动tigergraph:
gadmin start
rest导入(post)
http://192.168.1.234:9000/graph/work_graph
{
"vertices": {
"person": {
"id11": {
"id": {
"value": "sssssssss"
},
"gender": {
"value": "ttttttttttt"
}
}
}
}
}
gsql本地导入
USE GRAPH socialNet // v1.2
CREATE VERTEX engineer(PRIMARY_ID engineerId STRING, id STRING, locationId STRING, skillSet SET<INT>, skillList LIST<INT>, interestSet SET<STRING COMPRESS>, interestList LIST<STRING COMPRESS>)
CREATE VERTEX company(PRIMARY_ID clientId STRING, id STRING, country STRING)
CREATE UNDIRECTED EDGE worksFor(FROM engineer, TO company, startYear INT, startMonth INT, fullTime BOOL)
CREATE LOADING JOB loadMember FOR GRAPH socialNet {
DEFINE FILENAME f;
LOAD f
TO VERTEX engineer VALUES($0, $0, $1, _, _, SPLIT($3,"|"), SPLIT($3,"|") ),
TO TEMP_TABLE t2(id, skill) VALUES ($0, flatten($2,"|",1));
LOAD TEMP_TABLE t2
TO VERTEX engineer VALUES($0, _, _, $"skill", $"skill", _, _);
}
CREATE LOADING JOB loadCompany FOR GRAPH socialNet {
DEFINE FILENAME f;
LOAD f TO VERTEX company VALUES($0, $0, $1);
}
CREATE LOADING JOB loadMemberCompany FOR GRAPH socialNet {
DEFINE FILENAME f;
LOAD f TO EDGE worksFor VALUES($0, $1, $2, $3, $4);
}
RUN LOADING JOB loadMember USING f="./engineers"
RUN LOADING JOB loadCompany USING f="./companies"
RUN LOADING JOB loadMemberCompany USING f="./engineer_company"
java -jar /home/graph/tigergraph/tigergraph/dev/gdk/gsql/lib/gsql_client.jar ls
curl -X POST \
-H "Authorization: Basic dGlnZXJncmFwaDp0aWdlcmdyYXBo" \
-H "Cookie: {"TERMINAL_WIDTH":"270","CLIENT_PATH":"/home/graph","session":"3458564839407596.1421093678"}" \
-d "LS" \
http://127.0.0.1:8123/gsql/file
curl -X POST \
-H "Authorization: Basic dGlnZXJncmFwaDp0aWdlcmdyYXBo" \
-H ‘Cookie: {"TERMINAL_WIDTH":"270","CLIENT_PATH":"/home/graph","session":"3458564839407596.1421093678"}‘ \
-d "ls" http://127.0.0.1:8123/gsql/file
数据导入后文件存放路径:
/home/graph/tigergraph-2.2.4/tigergraph/gstore/0/part/.mv/21/1547711697574/vertex.bin
链接:
https://docs.tigergraph.com/admin/admin-guide/installation-and-configuration/installation-guide
查看当前模块状态:
gadmin status
gsql --version
清空数据的三种方法:
只是清空数据(只能用于超级用户):
gsql CLEAR GRAPH STORE -HARD
清空数据和表结构(和drop all效果一样,会停掉resttpp,gse,gpe模块):
gsql --reset
删除数据和表结构(报告顶点、边、图)会停掉resttpp,gse,gpe模块:
gsql DROP ALL
关闭tigergraph:
gadmin stop
启动tigergraph:
gadmin start
查看tiger所有配置:
gadmin --dump-config
列出tiger所有日志文件路径:
gadmin log
查看tiger的kakfa的主题:
cd /home/tiger/tigergraph/kafka/bin/
./kafka-topics.sh --zookeeper localhost:19999 --describe
Topic:deltaQ_GPE_1Q PartitionCount:1 ReplicationFactor:1 Configs:
Topic: deltaQ_GPE_1Q Partition: 0 Leader: 1 Replicas: 1 Isr: 1
Topic:post_log_Q PartitionCount:1 ReplicationFactor:1 Configs:
Topic: post_log_Q Partition: 0 Leader: 1 Replicas: 1 Isr: 1
Topic:post_log_Q_RESTPP_1Q PartitionCount:1 ReplicationFactor:1 Configs:
Topic: post_log_Q_RESTPP_1Q Partition: 0 Leader: 1 Replicas: 1 Isr: 1
默认有三个topic(都是1个分区):
deltaQ_GPE_1Q
post_log_Q
post_log_Q_RESTPP_1Q
查看tiger的kafka的有哪些消费组:
cd /home/tiger/tigergraph/kafka/bin/
./kafka-consumer-groups.sh --zookeeper localhost:19999 --list
查看kafka的topic的偏离值:
cd /home/tiger/tigergraph/kafka/bin/
./kafka-run-class.sh kafka.tools.GetOffsetShell --broker-list localhost:30002 --topic deltaQ_GPE_1Q
./kafka-run-class.sh kafka.tools.GetOffsetShell --broker-list localhost:30002 --topic post_log_Q
./kafka-run-class.sh kafka.tools.GetOffsetShell --broker-list localhost:30002 --topic post_log_Q_RESTPP_1Q
经过多轮数据接入性能测试表明(实际导入的数据是6406889 ,topic的偏离值deltaQ_GPE_1Q:0:6406889):
官方的restpp接口post数据的导入方式,是通过deltaQ_GPE_1Q这个kafka的topic进行缓冲的。
kafka消息格式是字节流
消费kafka:
/home/tiger/tigergraph/kafka/bin/kafka-console-consumer.sh \
--bootstrap-server localhost:30002 \
--topic deltaQ_GPE_1Q \
--consumer-property group.id=geektcp
如果从头开始消费,增加如下参数:
//--fromeginning
VERTEX<person> p1;
startpoint = to_vertex("c4", "company");
(必须在累加器前面定义,参考C++的语法规则):
TYPEDEF TUPLE <VERTEX v, EDGE e> CHILD;
OrAccum @@found = false;
OrAccum @notSeen = true;
ListAccum<VERTEX> @childsVetex;
ListAccum<EDGE> @childsEdges;
(和声明分开):
startpoint = to_vertex("c4", "company");
跟arangodb的filter不同,这不是一个通用过滤器,只能在伴随neighbors和neighborAttribute这两个函数时过滤,而且只能在ACCUM语句块中使用:
s.neighbors().filter(true),
v.@diffCountry += v.neighborAttribute("worksFor", "company", "id")
.filter(v.locationId != company.country),
JSONARRAY转LIST
JSONARRAY arr;
arr = filterObject.getJsonArray("filterEdgeType");
j = 0;
WHILE(j<arr.size()) DO
@@list += arr.getString(j);
j+=1;
END;
ACCUM关键字
用于一阶累加
POST-ACCUM关键字
用于二阶累加
一个语法糖(官方文档没有明确解释,但有这个用法,见《运算符、函数和表达式->表达式声明》):
二阶累加要调用一阶累加的变量时,加单引号即可。例如
V = SELECT s
FROM Start:s -(:e)-> :t
ACCUM t.@received_score += s.@score/(s.outdegree() +1),
t.@test += 2,
test1 = s.@score
POST-ACCUM// s.@score = (1.0-damping) + damping * s.@received_score,
s.@received_score = 0,
s.@score = 11,
@@maxDiff += abs(s.@score - s.@score‘),
test2 = s.@score‘,
test3 = s.@score;
POST-ACCUM 里面的s.@score‘ 其实就是 ACCUM的变量s.@score
SELECT或UPDATE语句的正文内不允许使用全局累加器赋值
在ACCUM或POST-ACCUM子句中允许对全局变量赋值,但只有退出该子句后该值才生效。 因此,如果同一全局变量有多个赋值语句,则只有最后一个赋值语句才会生效。
ACCUM子句中不允许使用顶点属性赋值。 但是,允许对边属性赋值。 这是因为ACCUM子句是由边的集合迭代而来。
使用通用的VERTEX作为元素类型的任何累加器都不能由LOADACCUM()初始化。
影响是:tiger支持基础条件的复杂组合过滤条件。不过这些复杂条件要写死在存储过程里面。
官方:NOT运算符不能与.type属性选择器结合使用。 要检查边或顶点类型是否不等于给定类型,请使用!=运算符
tiger的select语句其实本质上用C++语言,通过多线程,广度优先方式进行遍历。
这就是为什么全路径查询时,如果有多条路径时,结果集存在逆序。
CREATE QUERY expand_filter_pro3 (String filterStr) FOR GRAPH work_graph {
JSONOBJECT filter_object,filter_origin, filter_tmp;
INT filter_depth, filter_nodeSize, filter_sum, i, j;
BOOL filter_direction, filter_enable;
JSONARRAY filter_edges,filter_nodes,filter_origins, filter_collections,filter_value;
STRING filter_expression, filter_key;
MinAccum<INT> @accum_dis;
OrAccum @accum_visited;
ListAccum<VERTEX> @accum_path;
ListAccum<STRING> @accum_type;
OrAccum @@accum_on;
SetAccum<STRING> @@accum_edges;
SetAccum<STRING> @@accum_nodes;
ListAccum<VERTEX> @@accum_origins;
SetAccum<EDGE> @@accum_edges_result;
MapAccum<STRING, ListAccum<STRING>> @@filter_map;
ListAccum<STRING> @@filter_value;
filter_object = parse_json_object(filterStr);
filter_origins = filter_object.getJsonArray("origins");
filter_nodes = filter_object.getJsonArray("nodes");
filter_edges = filter_object.getJsonArray("edges");
filter_direction = filter_object.getBool("direction");
filter_enable = filter_object.getBool("enable");
filter_depth = filter_object.getInt("depth");
filter_nodeSize = filter_object.getInt("nodeSize");
filter_sum = filter_object.getInt("sum");
filter_expression = filter_object.getString("expression");
filter_collections = filter_object.getJsonArray("collection");
j = 0;
WHILE(j < filter_origins.size()) DO
filter_origin = filter_origins.getJsonObject(j);
@@accum_origins += to_vertex(filter_origin.getString("id"), filter_origin.getString("type"));
j += 1;
END;
j = 0;
WHILE(j < filter_edges.size()) DO
@@accum_edges += filter_edges.getString(j);
j += 1;
END;
j = 0;
WHILE(j < filter_nodes.size()) DO
@@accum_nodes += filter_nodes.getString(j);
j += 1;
END;
j = 0;
WHILE(j < filter_collections.size()) DO
filter_tmp = filter_collections.getJsonObject(j);
filter_key = filter_tmp.getString("key");
filter_value = filter_tmp.getJsonarray("value");
PRINT filter_value;
i=0;
WHILE(i < filter_value.size()) DO
@@filter_value += filter_value.getString(i);
PRINT @@filter_value;
i+=1;
END;
@@filter_map += (filter_key -> @@filter_value);
j += 1;
END;
PRINT @@filter_map;
//PRINT @@filter_map.get("company");
source = { @@accum_origins };
result = { @@accum_origins };
IF(filter_enable) THEN
source = SELECT s
FROM source:s - (:e) -> :t
ACCUM s.@accum_visited += true,
s.@accum_dis = 0,
s.@accum_path = s;
WHILE( source.size()>0 AND result.size() < filter_nodeSize) LIMIT filter_depth DO
source = SELECT t
FROM source:s -(:e)-> :t
WHERE t.@accum_visited == false
AND @@accum_edges.contains(e.type)
AND @@accum_nodes.contains(t.id)
AND s.@accum_dis < filter_sum
//AND t.city != "dongguan" OR NOT( @@filter_map.get("company").contains(t.city) AND t.sex == "female")
AND evaluate(filter_expression)
ACCUM t.@accum_dis += s.@accum_dis + 1,
t.@accum_path = s.@accum_path + [t],
t.@accum_visited += true,
@@accum_on += s.quadrant > 4,
@@accum_edges_result += e;
//HAVING t.city != "dongguan"
// OR ( t.city == "dongguan" AND t.sex == "female");
result = result UNION source;
END;
PRINT result as vertices;
PRINT @@accum_edges_result AS edges;
END;
}
CREATE QUERY shortest_path(VERTEX startpoint, VERTEX endpoint, INT depth = 5) FOR GRAPH work_graph {
OrAccum @@found = false;
OrAccum @notSeen = true;
ListAccum<VERTEX> @childsVetex;
ListAccum<EDGE> @childsEdges;
ListAccum<VERTEX> @@vertexes;
ListAccum<EDGE> @@edges;
MapAccum<VERTEX, ListAccum<VERTEX>> @@mapVertex;
MapAccum<VERTEX, ListAccum<EDGE>> @@mapEdge;
ListAccum<VERTEX> @@tmppoint;
ListAccum<VERTEX> @@tmpkeys;
String msg;
Start = { startpoint };
result = SELECT v
FROM Start:v
ACCUM v.@notSeen = false;
WHILE NOT @@found LIMIT depth DO
Start = SELECT t
FROM Start:s - (:e) -> :t
WHERE t.@notSeen
ACCUM t.@notSeen = false,
t.@childsVetex += s,
t.@childsEdges += e,
if t == endpoint THEN
@@found += true
END
POST-ACCUM
@@mapVertex += (t -> t.@childsVetex),
@@mapEdge += (t -> t.@childsEdges);
END;
@@tmppoint += endpoint;
@@vertexes += endpoint;
WHILE true LIMIT 10 DO
@@tmpkeys.clear();
FOREACH key in @@tmppoint DO
IF(@@mapVertex.containsKey(key)) THEN
@@tmpkeys += @@mapVertex.get(key);
@@vertexes += @@mapVertex.get(key);
@@edges += @@mapEdge.get(key);
END;
END;
@@tmppoint = @@tmpkeys;
IF(@@mapVertex.containsKey(startpoint)) THEN
BREAK;
END;
END;
IF @@found THEN
PRINT @@edges;
PRINT @@vertexes;
ELSE
msg = "Can‘t find shortest path within max depth";
PRINT msg ;
PRINT depth;
END;
}
CREATE QUERY full_path() FOR GRAPH work_graph {
VERTEX startpoint;
VERTEX endpoint;
INT graph_depth, result_depth;
OrAccum @@found_start = FALSE;
OrAccum @@found_end = FALSE;
AndAccum @@found = TRUE;
OrAccum @notSeen = TRUE;
SetAccum<VERTEX> @nextVetexes;
SetAccum<EDGE> @nextEdges;
SetAccum<VERTEX> @lastVetexes;
SetAccum<EDGE> @lastEdges;
SetAccum<VERTEX> @@vertexes;
SetAccum<EDGE> @@edges;
MapAccum<VERTEX, SetAccum<VERTEX>> @@mapVertex;
MapAccum<VERTEX, SetAccum<EDGE>> @@mapEdge;
MapAccum<VERTEX, SetAccum<VERTEX>> @@mapVertexReverse;
MapAccum<VERTEX, SetAccum<EDGE>> @@mapEdgeReverse;
SetAccum<VERTEX> @@tmppoint;
SetAccum<VERTEX> @@tmpkeys;
SetAccum<VERTEX> @@startlist;
SetAccum<VERTEX> @@endlist;
graph_depth = 6;
result_depth = graph_depth * 2;
startpoint = to_vertex("p7", "persons");
endpoint = to_vertex("s4", "skill");
@@startlist += startpoint;
@@endlist += endpoint;
Start (ANY) = {@@startlist};
result = SELECT v
FROM Start:v
ACCUM v.@notSeen = FALSE;
WHILE TRUE LIMIT graph_depth DO
Start = SELECT t
FROM Start:s - (:e) -> :t
WHERE t.@notSeen // and e.type !="all_to_skill"
ACCUM t.@notSeen = FALSE,
t.@lastVetexes += s,
t.@lastEdges += e,
s.@nextVetexes += t,
s.@nextEdges += e
POST-ACCUM
@@mapVertex += (t -> t.@lastVetexes),
@@mapEdge += (t -> t.@lastEdges),
@@mapVertexReverse += (s -> s.@nextVetexes),
@@mapEdgeReverse += (s -> s.@nextEdges);
END;
@@tmppoint += endpoint;
@@vertexes += endpoint;
WHILE TRUE LIMIT result_depth DO
@@tmpkeys.clear();
FOREACH key in @@tmppoint DO
IF(@@mapVertex.containsKey(key)) THEN
@@tmpkeys += @@mapVertex.get(key);
@@vertexes += @@mapVertex.get(key);
@@edges += @@mapEdge.get(key);
END;
IF(@@mapVertexReverse.containsKey(key)) THEN
@@tmpkeys += @@mapVertexReverse.get(key);
@@vertexes += @@mapVertexReverse.get(key);
@@edges += @@mapEdgeReverse.get(key);
END;
END;
@@tmppoint = @@tmpkeys;
END;
FOREACH point in @@startlist DO
IF(@@vertexes.contains(point)) THEN
@@found_start += TRUE;
END;
END;
FOREACH point in @@endlist DO
IF(@@vertexes.contains(point)) THEN
@@found_end += TRUE;
END;
END;
@@found += @@found_start;
@@found += @@found_end;
PRINT @@found;
PRINT @@edges;
PRINT @@vertexes;
}
高级过滤
高级过滤条件只能在结果集里面进行
因为高级过滤条件里面包含不只一个条件,如果只是查出一部分结果就进行过滤,那么这个filter要执行多次,这其实是低效的。每次查出一部分结果,比如多层展开,遍历第一层就解析过滤条件然后过滤一次,第二层又来一遍,效率很低。
传统关系型数据库的过滤条件where语句不是结果集里过滤的,而是查询过程中过滤。相当于图查询的每层展开过滤。
传统关系型数据库的复杂嵌套sql是从结果集里面过滤的,相当于图查询的结果集过滤。
tiger的存储过程中解析json,不能直接一步到位,想这样是语法是对的,但无法正常编译:
JSONOBJECT json;
json = parse_json_object( jsonStr );
param = json.getJsonObject("param");
PRINT param.getJsonObject("options").getString("edges");
正确的写法:
JSONOBJECT options;
options = json.getJsonObject("options");
PRINT options.getString("direction");
这要看是否启用HA,如果没有,那么文件锁每个节点都存放对应分片。
如果有HA,假定副本是两个,那么每个分片会在两个节点存放。
tiger分为两个版本:Developer Edition 和 Enterprise Edition
开发版只支持单点模式,并且不支持版本升级;
企业版支持单点和集群模式
序号 |
功能 |
开发版 |
企业版 |
1 |
集群 |
否 |
是 |
2 |
多图 |
否 |
是 |
3 |
修改图 |
否 |
是 |
4 |
节点数无限 |
否 |
根据license |
5 |
高可用 |
否 |
是 |
6 |
权限认证 |
否 |
是 |
7 |
数据备份 |
否 |
是 |
8 |
|
|
|
9 |
|
|
|
10 |
|
|
|
11 |
|
|
|
12 |
|
|
|
13 |
|
|
|
14 |
|
|
|
15 |
|
|
|
16 |
|
|
|
17 |
|
|
|
18 |
|
|
|
19 |
|
|
|
20 |
|
|
|
下载地址:
http://dl.tigergraph.com/developer-edition/tigergraph-2.2.4-developer-patch.tar.gz
官方文档:
https://docs.tigergraph.com/
官方开发文档:
https://www.tigergraph.com/developers/?utm_campaign=2018%20Developer%20Edition&utm_source=hs_automation&utm_medium=email&utm_content=63627934&_hsenc=p2ANqtz-9WiQx4ts5owPLxV7gy_d4JLvaLUTpa3axLLBObsV6c00DSzrw_gV9txh6Gr1af1RBpIeqBag9VWrvl2rxHlTTAHWTdLA&_hsmi=63627934
GQL:
https://doc.tigergraph.com/GSQL-101.html?utm_campaign=2018%20Developer%20Edition&utm_source=hs_automation&utm_medium=email&utm_content=63627934&_hsenc=p2ANqtz-9WiQx4ts5owPLxV7gy_d4JLvaLUTpa3axLLBObsV6c00DSzrw_gV9txh6Gr1af1RBpIeqBag9VWrvl2rxHlTTAHWTdLA&_hsmi=63627934
用户手册:
https://doc.tigergraph.com/TigerGraph-Platform-Knowledge-Base-and-FAQs.html?utm_campaign=2018%20Developer%20Edition&utm_source=hs_automation&utm_medium=email&utm_content=63627934&_hsenc=p2ANqtz-9WiQx4ts5owPLxV7gy_d4JLvaLUTpa3axLLBObsV6c00DSzrw_gV9txh6Gr1af1RBpIeqBag9VWrvl2rxHlTTAHWTdLA&_hsmi=63627934
web使用手册:
https://doc.tigergraph.com/TigerGraph-Platform-Overview.html?utm_campaign=2018%20Developer%20Edition&utm_source=hs_automation&utm_medium=email&utm_content=63627934&_hsenc=p2ANqtz-9WiQx4ts5owPLxV7gy_d4JLvaLUTpa3axLLBObsV6c00DSzrw_gV9txh6Gr1af1RBpIeqBag9VWrvl2rxHlTTAHWTdLA&_hsmi=63627934
tigergraph基础架构:
https://doc.tigergraph.com/TigerGraph-Platform-Overview.html?utm_campaign=2018%20Developer%20Edition&utm_source=hs_automation&utm_medium=email&utm_content=63627934&_hsenc=p2ANqtz-9WiQx4ts5owPLxV7gy_d4JLvaLUTpa3axLLBObsV6c00DSzrw_gV9txh6Gr1af1RBpIeqBag9VWrvl2rxHlTTAHWTdLA&_hsmi=63627934
rest api:
https://docs.tigergraph.com/dev/restpp-api/restpp-requests
构建测试数据:
https://docs.tigergraph.com/dev/gsql-ref/querying/appendix/example-graphs#socialnet
Tigergraph高性能图数据库深入调研
标签:groupadd red img ase 执行sql ora member lib 机制