doris 1.2.1 query iceberg. it's mins level. need help to analyze 求助

[bossjie]

FE 3个节点, 16vcpu.64GBmem.80GBssd.2000eph
BE 4个节点, 16vcpu.64GBmem.664GBssd.2000eph
在master节点上 用mysql client, 创建catalog.
能顺利读取iceberg里的database和对应tables.
执行以下很小的表(dt是partition key). 第一次需要3 分多钟 ,第二次 第三次 会减少到几十秒.
doris集群到iceberg集群都是内网(万兆网卡). 延迟0.x 毫秒
查询时 我也看了机器 cpu/mem, 都没飙高
也尝试了iceberg里的其他database的table,也是巨慢.
查看了fe.log 也没error见着.
select count() from pda.npsjoincallanalyzeaggregatehive where dt>='2023-09-01'; //dt为partition key
+----------+
| count() |
+----------+
| 128522 |
+----------+
1 row in set (17.21 sec)

如果可以的话, 十分希望加vx聊. 1051271623

感谢

[LemonLiTree]

try upgrading to 2.0.2

[bossjie]

@LemonLiTree thanks for replying my ask.
https://www.infoq.cn/article/8su754vfukpw8gwjf24v
this blog says the speed to query iceberg is 3~10 faster than trino(presto).
and it's using 1.2.x as I believe. I believe it's not release version related and big companies keep using Doris, they definitely have a good performance when querying iceberg using old version.
Before upgrading version to 2.0.2, I'd like to learn what's happened in my scenerio. Thanks.