Skip site navigation (1) Skip section navigation (2)

Re: Lower Random Access Time vs RAID 0 / 1

From: Ron <rjpeace(at)earthlink(dot)net>
To: "Michael Ben-Nes" <miki(at)epoch(dot)co(dot)il>, "PostgreSQL Performance" <pgsql-performance(at)postgresql(dot)org>
Subject: Re: Lower Random Access Time vs RAID 0 / 1
Date: 2007-03-22 14:37:04
Message-ID: E1HUOPX-0004DE-Ts@elasmtp-junco.atl.sa.earthlink.net (view raw or flat)
Thread:
Lists: pgsql-performance
1= a better HD comparison resource can be found at www.storagereview.com
http://www.storagereview.com/comparison.html

You will find that storagereview has better information on any and 
all things HD than Tom's does.


2= DB servers work best with as many spindles as possible.  None of 
your example configurations is adequate; and any configuration with 
only 1 HD is a data loss / data corruption disaster waiting to happen.
In general, the more spindles the better with any DB.  The =minimum= 
should be at least 4 HD's =dedicated= to the DB.  OS HD's are 
independent and in addition to the 4+ DB HDs.


3= "heavy duty large DB with mostly reads and heavy write actions 
from time to time ( updates / huge transactions )." Does not have 
anywhere near the precision needed to adequately describe your needs 
in engineering terms.
How big a DB?
What % of the IO will be reads?  % writes?
How big is a "huge transaction"?
Exactly what is the primary use case of this server?
etc.  We need =numbers= if we are going to think about "speeds and 
feeds" and specify HW.


4= =seriously= consider HW RAID controllers like 3ware (AKA AMCC) or 
Areca. with BB IO caches.


You've got a lot more work ahead of you.
Ron


At 05:08 AM 3/22/2007, Michael Ben-Nes wrote:
>Hello
>
>I plan to buy a new development server and I wonder what will be the 
>best HD combination.
>
>I'm aware that "best combination" also relay on DB structure and usage.
>so lets assume, heavy duty large DB with mostly reads and heavy 
>write actions from time to time ( updates / huge transactions ).
>
>Here are the options:
>
>One very fast 10K RPM SATA Western Digital Raptor 150GB HD.
>   Pro: very low access time and generally 30% faster regarding mainstream HD.
>   Con: Expensive.
>
>2 mainstream 7.2K RPM SATA HD in RAID 0.
>   Pro: fast transfer rate.
>   Con: Access time is lowered as both HD has to sync for read / 
> write ( true ? ).
>
>2 mainstream 7.2K RPM SATA HD in RAID 1.
>   Pro: can access parallely different files in the same time ( true ? ).
>   Con: Slower at writing.
>
>Random access benchmark:
><http://www23.tomshardware.com/storage.html?modelx=33&model1=280&model2=675&chart=32>http://www23.tomshardware.com/storage.html?modelx=33&model1=280&model2=675&chart=32 
>
>
>Will be happy to hear recommendations and ideas.
>
>Thanks,
>Miki
>
>--
>--------------------------------------------------
>Michael Ben-Nes - Internet Consultant and  Director.
><http://www.epoch.co.il>http://www.epoch.co.il - weaving the Net.
>Cellular: 054-4848113
>--------------------------------------------------


In response to

pgsql-performance by date

Next:From: Michael FuhrDate: 2007-03-22 14:37:13
Subject: Re: Performance of count(*)
Previous:From: Mario WeilguniDate: 2007-03-22 14:36:32
Subject: Re: Performance of count(*)

Privacy Policy | About PostgreSQL
Copyright © 1996-2014 The PostgreSQL Global Development Group