Exadata, You Bully !
Yesterday I was reviewing a presentation and had an “aha” moment. In one case, results were being compared between technologies that were at about the same currency. In another case comparisons were being done between brand new spanking technology — and systems that were many years old.
This morning I again came face to face with this fallacy. Once again, on the front page of the Wall Street Journal, Oracle compares their technology in use at an anonymous European retailer at level x with IBM technology at level x-n (n is some integer, depending on how you count your releases).
The Oracle Exadata system in the comparison was much newer, larger, had over 3x the memory, was running a newer database version, and used specialized storage.
Do any claims in a situation such as this make any sense at all ?
Sort of like deciding to have a fistfight with your grandmother.
************************************************
The postings on this site solely reflect the personal views of the author and do not necessarily represent the views, positions, strategies or opinions of IBM or IBM management.
technorati tags: IBM, systems, performance,Exadata,595,Oracle,11g,10g,benchmark,Wall Street,Journal,XP24000,HP,database,AIX,hardware,Power,benchmarking
To be fair to Oracle they have to spend that enormous marketing budget on something – and since they don’t publish benchmarking results for Exadata it has to go on FUD and stories of anonymous wins against unknown systems!
flashdba
September 13, 2012 at 10:47 am
Ad must have been only in the print edition?
kevinclosson
September 13, 2012 at 12:28 pm
http://www.oracle.com/us/corporate/features/euro-retailer/index.html quote:
“Up to 15-20x faster queries across 2.8 billion records on count/sum queries”
Elisabeth,
Typical Exaggerdata.
Your readers need to question the business value of the occasional count(*) from sum(col1) scans against a single table. My Exadata Critical Thinking video (http://wp.me/p21zc-Xf) explains data flow dynamics in Exadata and why “proof points” like simple light scans are not indicative of what the platform can do. I urge your readers to dedicate the 1 hour it takes to view that series of video presentations.
Oracle 10g compared to Oracle 11g means there was time taken to performance a migration. Indeed, the migration included an endianess-jump so we know there was a lot of consulting involved. Since it was a migration to Exadata Oracle put their best and brightest on the SQL and we all know that SQL tuning is the first line of defense in RDBMS performance.
In short, this is an apples to bicycles comparison.
Typical, but sad nonetheless.
kevinclosson
September 13, 2012 at 2:10 pm