Hello *,
Long time ago I used VIEWs on my history database and when PostgreSQL
introduced table partitioning I was switching to it.
Currently I use Debian GNU/Linux 7.11 with PostgreSQL 9.1 (table space
and table partitioning) using an Adaptec 16-Channel Raid-1 Controller
with 16x 6 TByte SAS UltraStar (HGST) drives.
I created partitions of 100 years range and have currently 132 of them.
There are 8 base columns (1 data colum in english) and then for each
language used an additional colum. Some of the data where translated
into 27 languages but I have 56 non-english columns.
So in total I have currently 64 columns and arround 156mio rows.
Th etables are organised into
history_earth
history_earth_021_a
history_earth_020_a
history_earth_019_a
...
history_earth_124_b
...
"a" mean A.C. and "b" B.C.
While checking my Root-Servers I discovered, that my PostgreSQL Box has
a huge performance problem, because some table partitions became VERY
large (3 table partitions exceed already 1 TByte and 14 are 500-1000GB).
I like to make the partitions smaller, but the documentation say, you
should not dare to make several 1000 partitions..
Question:
What is with is today possibel with table partitioning?
And there is another thing:
The 16 "new" SAS Drives where sponsored, hence free for me including the
new Raid-1 Controller, but I do not like the idea, to have a singel
physical database of arround 40 TByte diskspace...
I can get much less expensive (not cheaper) SAS drives of 1 or 2 TByte
and since the Raid-1 Controller cost only 400, I can use several LOW
POWER machines (Mini-ITX) to accomplish the task because the webinterface
is anyway stored on a seperated server.
Any suggestions?
--
Michelle Konzack
00372-54541400
