4G row table?

From: george young <gry(at)ll(dot)mit(dot)edu>
To: pgsql-performance(at)postgresql(dot)org
Subject: 4G row table?
Date: 2002-12-19 19:10:58
Message-ID: 20021219141058.11d4b693.gry@ll.mit.edu
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

[linux, 700MHz athlon, 512MB RAM, 700GB 10kRPM SCSI HW RAID, postgresql 7.2]
We're setting up a DB of IC test data, which is very simple and regular, but large.
One project (we get three or four per year) has ~4 giga bits, each specified by
a few small integer values, e.g.:
Name Type Values
----------------------
wafer int 1-50
die_row int 2-7
die_col int 2-7
testtype string (~10 different short strings)
vdd int 0-25
bit_col int 0-127
bit_row int 0-511
value bit 0 or 1

with 4 billion(4e9) rows. I would guess to make wafer, die_row, etc. be of
type "char", probably testtype a char too with a separate testtype lookup table.
Even so, this will be a huge table.

Questions: How much overhead will there be in the table in addition to the
9 bytes of data I see? How big will the primary index on the first seven columns
be? Will this schema work at all?

Of course, we could pack 128 bits into an 8 byte "text" field (or should we use bit(128)?),
but lose some ease of use, especially for naive (but important) users.

Comments, suggestions?

-- George

--
I cannot think why the whole bed of the ocean is
not one solid mass of oysters, so prolific they seem. Ah,
I am wandering! Strange how the brain controls the brain!
-- Sherlock Holmes in "The Dying Detective"

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Josh Berkus 2002-12-19 19:15:20 Re: 4G row table?
Previous Message Manfred Koizar 2002-12-19 19:07:53 Re: 4G row table?