Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> comp.databases.theory -> Re: Nearest Common Ancestor Report (XDb1's $1000 Challenge)

Re: Nearest Common Ancestor Report (XDb1's $1000 Challenge)

From: Neo <neo55592_at_hotmail.com>
Date: 2 Jun 2004 17:09:24 -0700
Message-ID: <4b45d3ad.0406021609.299d6c01@posting.google.com>


Small Report Generation Summary (provided by Hugo)


Solution   Time(ms)  Platform            Notes
---------- --------  ----------          ----------------------------
RM's #1       14.3   1.3 Ghz PC          Apple-to-orange
RM's #2       11.0   1.3 Ghz PC          A step closer to apples
XDb1 4.4.7    16     1.3 Ghz PC          Unoptimized, debug version


Small Report Generation Summary (provided by Neo)



Solution Time(ms) Platform Notes
---------- --------  -----------------  -----------------------------
RM's #1      65.0    500 Mhz Server     Apple-to-orange
RM's #2      68.9    500 Mhz Server     A step closer to apples
RM's #4     152.13   500 Mhz Server     Another step closer to apples
XDb1 4.4.7   16      500 Mhz Server     Unoptimized, debug version
XDb1 4.5.7    1.632  500 Mhz Server     Optimized version
XDb1 4.5.9 6.561 233 MHz Pocket PC 32 MB

Large Report (28,940 rows) Generation Summary (provided by Neo)



Solution Time(sec) Platform Notes
---------- --------  -----------------  ----------------------------
RM's #1      15.2    500 Mhz Server     Apple-to-orange
XDb1 4.5.7    2.9    500 Mhz Server     Optimized version


Since RM Solution #2, didn't generate a class hierarchy report and didn't allow things with no or multiple names, etc... I have updated it to RM Solution#4 (see script below) which is another step closer to making an apple-to-apple comparision. The class hierarchy which had been stored in redundant tables T_Classes, T_ClassOfThings and T_ClassHierarchy have been dropped and the data moved to T_hierarchies, thus allowing a report on the class hierarchy also. In addition, the name attribute of things was moved to T_attributes_char, thus allowing a thing with no name or multiple names. However, T_attributes_char will have redundant data if a thing's property has multiple values. The next RM Sol should normalize by adding T_AttribChar_Value. As one can see from the measurements above, RM's solution gets slower as it is begins to approach the level of genericness and normalization in XDb1 (which is normalized down to atomic symbols).

If someone can also run the below script on SQL Server for PocketPC, I would be interested in comparing its performance.

Received on Wed Jun 02 2004 - 19:09:24 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US