结果集行的平均数

Question

我有一个带宽利用率数据表。每行都有主机名，niccardname，利用率和时间戳。在每个主机的最大时间戳处，可能有不同的NIC卡。 因此对于每个主机，我希望在最大时间戳下不同NIC卡的平均利用率。

下面是我的表结构，插入和查询 -

CREATE TABLE bandwith_utilization 
(
    id integer NOT NULL, 
    hostname character varying(255), 
    "timestamp" bigint, 
    niccardname character varying(255), 
    percentageutilization integer, 
    CONSTRAINT bandwidth_utilization_pkey PRIMARY KEY (id) 
) 
WITH (
    OIDS=FALSE 
); 
ALTER TABLE bandwith_utilization 
    OWNER TO postgres; 


INSERT INTO bandwith_utilization 
VALUES (1,'host1','111111','nic1',40); 
INSERT INTO bandwith_utilization 
VALUES (2,'host1','111112','nic1',50); 
INSERT INTO bandwith_utilization 
VALUES (3,'host1','111113','nic1',50); 
INSERT INTO bandwith_utilization 
VALUES (4,'host1','111113','nic2',70); 

INSERT INTO bandwith_utilization 
VALUES (5,'host2','111111','nic1',80); 
INSERT INTO bandwith_utilization 
VALUES (6,'host2','111112','nic1',20); 
INSERT INTO bandwith_utilization 
VALUES (7,'host2','111112','nic2',30); 

INSERT INTO bandwith_utilization 
VALUES (8,'host3','111115','nic1',10); 

所以这是插入后，我的表 -

id hostname timestamp niccardname  percentageutilization 
------------------------------------------------------------------ 
1; "host1"; 111111;  "nic1";   40 
2; "host1"; 111112;  "nic1";   50 
3; "host1"; 111113;  "nic1";   50 
4; "host1"; 111113;  "nic2";   70 

5; "host2"; 111111;  "nic1";   80 
6; "host2"; 111112;  "nic1";   20 
7; "host2"; 111112;  "nic2";   30 

8; "host3"; 111115;  "nic1";   10 

我有一个查询这给我的输出在最大时间戳主机名 -

select hostname, timestamp, niccardname, percentageutilization 
from report.bandwith_utilization 
where timestamp = (select max(timestamp) 
        from report.bandwith_utilization nwUtil 
        where nwUtil.hostname = report.bandwith_utilization.hostname 
        ) ; 

以上查询的输出是 -

"host1"; 111113; "nic1"; 50 
"host1"; 111113; "nic2"; 70 

"host2"; 111112; "nic1"; 20 
"host2"; 111112; "nic2"; 30 

"host3"; 111115; "nic1"; 10 

所以现在我的预期输出是每个主机不同NIC卡的平均利用率。即

"host1"; 111113; "nic1"; 60 
"host2"; 111112; "nic1"; 25 
"host3"; 111115; "nic1"; 10 

我如何才能找到最后的平均输出与我在上面提到的相同的查询？

Answer 1

应该是一个AVG（）和GROUP BY

select hostname,timestamp,min(niccardname), avg(percentageutilization) 
from report.bandwith_utilization 
where (timestamp,hostname, niccardname) in (select max(timestamp) ,hostname, niccardname 
from report.bandwith_utilization nwUtil 
where nwUtil.hostname= report.bandwith_utilization.hostname 
group by hostname, niccardname 
) 
group by hostname,timestamp 
order by hostname,timestamp 

Answer 2

下面是一个好办法，最大时间戳来获得行：

select u.* 
from (select u.*, 
      rank() over (partition by hostname order by timestamp desc) as seqnum 
     from report.bandwith_utilization u 
    ) u 
where seqnum = 1; 

现在，你可以得到你想要的想：

select u.hostname, u.timestamp, avg(percentageutilization) 
from (select u.*, 
      rank() over (partition by hostname order by timestamp desc) as seqnum 
     from report.bandwith_utilization u 
    ) u 
where seqnum = 1 
group by u.hostname, u.timestamp; 

它没有意义的（对我）包括在结果集中niccardname。如果你想要一个值，你可以使用min(niccardname)，如果你想让它们都在数组中，你可以使用min(niccardname)或array_agg(niccardname)。