ntile 是你的朋友. (Oracle 與 SQL 都有, 以下是用 Oracle 的 LiveSQL 做範例)

我向來土法練鋼，遵古法製，沒有奇門遁甲～

select
	min(gts) AS startts,
	max(gts) AS endts,
	min(val) AS minval,
	max(val) AS maxval	
from
	it201120,
	(SELECT @stime:= (select min(gts) from it201120)) AS a,		
	(select @nSecond := (select (UNIX_TIMESTAMP((select max(gts) from it201120)) - UNIX_TIMESTAMP(@stime))/99)) AS b
group by round((UNIX_TIMESTAMP(gts) - UNIX_TIMESTAMP(@stime))/@nSecond,0);

https://dbfiddle.uk/?rdbms=mysql_8.0&fiddle=20be485c98da8227418363292f59873a

資料表結構

CREATE TABLE IT201120 (
  id int PRIMARY KEY AUTO_INCREMENT,
  gts timestamp,
  val int NOT NULL
)

隨機產生資料

SET @@cte_max_recursion_depth  = 100000;
SET @Today = CURDATE() + interval 0 second;

-- 隨機產生10萬筆時間與數值資料
INSERT INTO IT201120 (gts, val)
WITH RECURSIVE GenerateRandomTime AS (
  SELECT @Today AS DateTimeField, 1 AS RowNum
  UNION ALL
  SELECT @Today + interval ceil(RAND() * 86400) second AS DateTimeField, RowNum + 1
  FROM GenerateRandomTime
  WHERE RowNum < 1e5
)
SELECT DateTimeField, (RAND() * 1e6)
FROM GenerateRandomTime;

取得最大與最小時間

-- 最大時間與最小時間
SELECT MIN(gts), 
       MAX(gts)
FROM IT201120

依照題意排序後分組，取得題目的結果

-- 依照資料筆數分成100組
-- 取當組別的開始與結束時間(gts)，以及該組別的最大與最小值(val)
SET @rowPerGroup = (SELECT COUNT(*) / 100 FROM IT201120);

SELECT groupIndex, MIN(gts), MAX(gts), MIN(val), MAX(val)
FROM (
      SELECT ceil(ROW_NUMBER() OVER (ORDER BY gts) / @rowPerGroup ) AS groupIndex, gts, val
      FROM IT201120
) AS GroupResult
GROUP BY groupIndex

我承認臨時抱佛腳，OPEN BOOK學POSTGRESQL
請砍小力一點

既然是86400秒內隨意分組，那我就切成100組，以864秒為一組來切

Schema (PostgreSQL v13 (Beta))

create table it201120 (
  id int generated always as identity
, gts timestamp not null
, val int not null
);

-- 輸入10萬筆

insert into it201120(gts, val)
select 'yesterday'::timestamp + interval '1 second' * ceil(random() * 86400)
     , floor(random() * 1e6)
  from generate_series(1, 1e5);
  
select min(gts)
     , max(gts)
  from it201120;

Query #1

with five_min_intervals as (
  select 
    (select min(gts)::date from it201120) + ( n    || ' seconds')::interval start_time,
    (select min(gts)::date from it201120) + ((n+864) || ' seconds')::interval end_time
  from generate_series(0, (24*60*60), 864) n
)
select min(m.gts), max(m.gts),min(m.val),max(m.val)
from it201120 m
right join five_min_intervals f 
        on m.gts >= f.start_time and m.gts < f.end_time
group by f.start_time, f.end_time
order by f.start_time;

View on DB Fiddle