Re: {Spam?} Re: [HP3000-L] Howmessy

At 02:36 PM 2008-02-08, Roy Brown wrote:
If accessed heavily, yes.

Increasing capacity as I recommended should reduce %secondaries as well as max. blocks.




Not true, surely? Every DBGET is nearly one-and-a-half DBGETs, compared with a well-distributed master. That will slow things down, not just for this master, but for every detail access using it for the chain head.

And it's certainly true that every pointer is crossing a block boundary, hence the Ineff of 100%. So, yes, up the block factor. How big is a record?

Your analysis doesn't show what the Search Filed and Type is; it's not binary, is it?

But under 'Master Dataset Solutions', the HowMessy manual says:

"If secondaries are over 30% and inefficient pointers are over 50%, the dataset is either too full or not hashing properly. Increase capacity to a higher odd number, or change the data type and format of the search field.

Increasing block factor should reduce inefficient pointers.!

Both these prescriptions apply to the figures you give.

Roy

At 01:20 PM 2008-02-08, Jim Phillips wrote:
So, I ran Howmessy on the main data base here and the very first data set looks very suspicious. The report for this manual master shows 78% full (19381 entries, 24859 capacity), 30.3% secondaries, maximum blocks=190, blocking factor=1, maximum chain=7, average chain=1.43, standard deviation=.70, expanded blocks=1.43, average blocks=2.30, elongation=1.60 and, are you ready for this, inefficient pointers=100%!

My first thought was to increase the blocking factor (my thinking is with a blocking factor of 1, every pointer must cross a block boundary). I could also increase the capacity somewhat to get the load factor down to 70% or so.

What does this august group recommend?

Jim


---------------------------------
Looking for last minute shopping deals? Find them fast with Yahoo! Search.

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *



--
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.5.516 / Virus Database: 269.19.21/1265 - Release Date: 2008-02-07 11:17 AM

------------------------------------------------------------------------
-------------------------
Gilles Schipper
GSA Inc.
HP System Administration Specialists
300 John Street, Box 87651 Thornhill, ON Canada L3T 7R4
Voice: 905.889.3000 Fax: 905.889.3001
email: gilles@xxxxxxxxxx web: http://www.gsainc.com
------------------------------------------------------------------------
-------------------------

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

--
Roy Brown 'Have nothing in your houses that you do not know to be
Kelmscott Ltd useful, or believe to be beautiful' William Morris

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *



No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.5.516 / Virus Database: 269.19.21/1265 - Release Date: 2008-02-07 11:17 AM

-------------------------------------------------------------------------------------------------
Gilles Schipper
GSA Inc.
HP System Administration Specialists
300 John Street, Box 87651 Thornhill, ON Canada L3T 7R4
Voice: 905.889.3000 Fax: 905.889.3001
email: gilles@xxxxxxxxxx web: http://www.gsainc.com
-------------------------------------------------------------------------------------------------

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

.

Relevant Pages

  • Re: Howmessy
    ... That will slow things down, not just for this master, but for every detail access using it for the chain head. ... My first thought was to increase the blocking factor (my thinking is with a blocking factor of 1, every pointer must cross a block boundary). ... To join/leave the list, search archives, change list settings, * ...
    (comp.sys.hp.mpe)
  • Re: Howmessy
    ... TYPE COUNT CAPACITY COUNT LENGTH FACTOR ... To join/leave the list, search archives, change list settings, * ...
    (comp.sys.hp.mpe)
  • Re: Howmessy
    ... TYPE COUNT CAPACITY COUNT LENGTH FACTOR ... To join/leave the list, search archives, change list settings, * ...
    (comp.sys.hp.mpe)
  • Re: Howmessy
    ... TYPE COUNT CAPACITY COUNT LENGTH FACTOR ... To join/leave the list, search archives, change list settings, * ...
    (comp.sys.hp.mpe)
  • Re: Howmessy
    ... Before you change the capacity, could you at least tell us the size of the ... IMAGE record and maybe show a listf,2 of the dataset in question? ... To join/leave the list, search archives, change list settings, * ...
    (comp.sys.hp.mpe)