RE: Flash technology based HDD will it make significant difference for OLTP applications?
Date: Mon, 7 Jan 2008 14:33:41 -0500
May I speak from actual experience? The device you have in mind is a darned nice one and it will stand up to the specs (BTUT), but don't expect much improvement at the Oracle level for the expense. Yes Memory drives like this have come down in cost significantly, but so have their more primitive cousins and even more so. The expense of a larger cache in the disk subsystem far out weights the cost of a RAM drive and most disk subsystems currently provide near the same response times, mainly because your dealing with the cache not the actual drive. Again you get into that rat trap of poorly written SQL that holds things up when they should not. I believe it was Cary Millsap who once wrote a paper of the excessive cost of Logical IO's over Physical IO's and I'm afraid that I've seen far more of those problems than one would have hoped for. The BIG problem is recursive SQL that pounds one very small table or index into oblivion. And adding one of these RAM drives into the mix only makes matters worse, not better.
Now if you're going to create a standalone database server that isn't going to have access to a disk array, that's a bird of a different flock.
Dick Goulet / Capgemini
North America P&C / East Business Unit
Senior Oracle DBA / Hosting
Office: 508.573.1978 / Mobile: 508.742.5795 / www.capgemini.com Fax: 508.229.2019 / Email: richard.goulet_at_capgemini.com 45 Bartlett St. / Marlborough, MA 01752
Together: the Collaborative Business Experience
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Jurijs Velikanovs Sent: Sunday, January 06, 2008 4:35 PM
Subject: Flash technology based HDD will it make significant difference for OLTP applications?
Hello everyone and Happy new year to you,
Just wonder if you can give your advice if the devices as following
will make any difference in the context for Oracle RDBMS performance
in the nearest time?
From my point of view:
- "Access time <0.02 ms" at least 30 times faster then today's HDD-s
capable to achieve.It gives significant improvement for some profile Physical IO-s.
- Current price 5k for 300GB might be acceptable for some systems even
today. This parameter will go down within 1-2 years significantly.
- Max read/write operation count is limited. In spite of that to use
first point some organizations might choose to change HDD on regular basis (once a year/months) to improve throughput for some applications
- It will not solve all problems but introduce significant
improvements in some areas.
PS I am sorry if this topic has been discussed already, please point me to the thread then.
Thank you in advance,
-- http://www.freelists.org/webpage/oracle-l This message contains information that may be privileged or confidential and is the property of the Capgemini Group. It is intended only for the person to whom it is addressed. If you are not the intended recipient, you are not authorized to read, print, retain, copy, disseminate, distribute, or use this message or any part thereof. If you receive this message in error, please notify the sender immediately and delete all copies of this message. -- http://www.freelists.org/webpage/oracle-lReceived on Mon Jan 07 2008 - 13:33:41 CST