内网查询网络耗时太大 ResultDeliverTime

内网查询单表8e的数据,获取80w行,耗时10s
CREATE TABLE forward_return_table (
instrument_id varchar(64) NOT NULL COMMENT “”,
trade_time datetime NOT NULL COMMENT “”,
factor_id varchar(64) NOT NULL COMMENT “”,
value double NOT NULL COMMENT “”,
update_time datetime NOT NULL DEFAULT CURRENT_TIMESTAMP COMMENT “”
) ENGINE=OLAP
PRIMARY KEY(instrument_id, trade_time, factor_id)
PARTITION BY date_trunc(‘day’, trade_time)
DISTRIBUTED BY HASH(instrument_id, factor_id)
ORDER BY(instrument_id, trade_time)
PROPERTIES (
“compression” = “LZ4”,
“enable_persistent_index” = “true”,
“fast_schema_evolution” = “true”,
“replicated_storage” = “true”,
“replication_num” = “1”
);

【是否存算分离】否
【StarRocks版本】3.3.9
【集群规模】1fe 1be docker compose 部署
【机器信息】48C/1000G/万兆 , SSD 100T(read: 7000MB/s, write: 2000MB/s)
【附件】
profile.txt (13.4 KB)

连接数据库时序列化和和网络开销太大了,新版的arrow接口应该能解决。目前可以先远程写到parquet文件中再下载回来解析,曲线救国