sql - How can I fetch the last N rows, WITHOUT ordering the table -

i have tables multiple million rows , need fetch last rows of specific id's

for example last row has device_id = 123 , last row has device_id = 1234

because tables huge , ordering takes time, possible select last 200 without ordering table , order 200 , fetch rows need.

how that?

thank in advance help!

update

my postgresql version 9.2.1

sample data:

time                      device_id         data       data .... "2013-03-23 03:58:00-04" | "001ec60018e36" | 66819.59 | 4.203 "2013-03-23 03:59:00-04" | "001ec60018e37" | 64277.22 | 4.234 "2013-03-23 03:59:00-04" | "001ec60018e23" | 46841.75 | 2.141 "2013-03-23 04:00:00-04" | "001ec60018e21" | 69697.38 | 4.906 "2013-03-23 04:00:00-04" | "001ec600192524"| 69452.69 | 2.844 "2013-03-23 04:01:00-04" | "001ec60018e21" | 69697.47 | 5.156 .... 

see sqlfiddle of data

so if device_id = 001ec60018e21 want recent row device_id. grantee last row device_id row want, may or may not last row of table.

the general way "last" row each device_id looks this.

select * table1  inner join (select device_id, max(time) max_time             table1             group device_id) t2    on table1.device_id = t2.device_id   , table1.time = t2.max_time; 

getting "last" 200 device_id numbers without using order isn't practical, it's not clear why might want in first place. if 200 arbitrary number, can better performance taking subset of table that's based on arbitrary time instead.

select * table1  inner join (select device_id, max(time) max_time             table1             time > '2013-03-23 12:03'             group device_id) t2 on table1.device_id = t2.device_id , table1.time = t2.max_time;