Hi -
I'm running Postgres v8.3 on Fedora 10 (home machine - x86_64
architecture). After realizing that the storage requirements of one of
my databases will exceed 4Tb, I wanted to see if anyone had any
suggestions as to hardware setup that works well with Postgres running
on Linux. I have partitioned most of the database, so older data can
go on slower devices with only 400-500Gb needed on faster devices.
Redundancy is a requirement. General database usage is generally low,
with burst input. For the curious, here's the general profile:
a) ~19,000,000 INSERT statements/ day (190 INSERTs/ commit blocks)
for a total of ~10 Gb INSERT/day
b) INSERTs occur sequentially (daily batch)
c) Handful of SELECT statements run per day to generate a few
dozen reports, but these are not resource-intensive.
d) Only 3-5 database users configured.
Initial thought: Use an external multi-bay eSATA case and inserting
some 7200RPM+, 32Mb cache 1T+ SATA drives running RAID 5.
A few specific questions:
1) Which components generally lead to bottlenecks (controller
card, individual HD cache, HD rotational speed, other???)
2) Are there better solutions (broad question in terms of price/
data availability)? Budget: $500 - $750 for the storage medium. Better
question: Which setup was worked well with a Postgres implementation
running on a similar budget?
3) If using a similar setup, are there any particular
implementation issues that are not relatively obvious?
Thanks for the input (even if not Postgres-specifc, I thought this
might be of interest to hobby database admins like myself running
Postgres implementations).
