Re: OT hardware recommend

From: "Mike Sofen" <msofen(at)runbox(dot)com>
To: <pgsql-general(at)postgresql(dot)org>
Subject: Re: OT hardware recommend
Date: 2016-06-18 21:06:28
Message-ID: 05a401d1c9a5$494b3f30$dbe1bd90$@runbox.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

From: Adam Brusselback
Sent: Saturday, June 18, 2016 1:26 PM
Subject: Re: [GENERAL] OT hardware recommend

It's really amazing how much solid state drives transferred the database bottleneck away from disk.

Adam – so very true. We used to spend ungodly amounts of time/money/effort to mitigate disk performance limitations. It is almost MAGIC what SSDs do now.

Real numbers don’t lie: 5400 rpm disks can muster no more than 65 IOPS (7200s get ~90-100, 10k get 140-150). So:

15 x 65 = 975 IOPS (aka boohoo)

Using the AS SSD Benchmark, the Samsung 480gb m2 850 EVO in my core i7 laptop measures (IOPS then MB/s):

Random 4k blocks: 7,235 iops read, 14,012 iops

Random 4K-64Threads: 97,743 iops read, 68,864 iops write

Random 512B: 14,380 iops read, 19,858 iops write (db comparison here)

MB/s:

Sequential: 500 MB/s read, 449 MB/s write

Random 4K: 28.26 MB/s read, 54.74 MB/s write

4K-64Threads: 381.81 MB/s read, 269.00 MB/s write (this is closer to what db access looks like).

Access Times: 0.070 ms read, 0.050 ms write

Thusly,

1 x SSD = 14.75 times faster than a 15 drive array on reads, and 20 times faster on writes.

Like everyone else has said, just buy a 1 TB Samsung EVO 850 for $300 (USD) and call it a day. :)

Mike

In response to

Browse pgsql-general by date

  From Date Subject
Next Message Tim Smith 2016-06-19 08:58:58 Help needed structuring Postgresql correlation query
Previous Message Adam Brusselback 2016-06-18 20:25:32 Re: OT hardware recommend