Dynamic Files vs Static Hashed - Performance Issues

Dynamic Files vs Static Hashed - Performance Issues

Post by Barry Gausde » Wed, 09 Jun 1999 04:00:00



I don't want to provoke another monster thread on this. Can anyone point me
at a documentation/discussion regarding the performance penalties associated
with dynamic files in Universe.

I don't necessarily want a discussion of linear hashing algorithms or file
structure, I'm looking for (ideally) numbers showing comparitive
read/update/select times for dynamic vs static, particularly for files
exceeding 1GB/1M records.

Barry.

 
 
 

Dynamic Files vs Static Hashed - Performance Issues

Post by Anthony Dzikiewic » Sat, 12 Jun 1999 04:00:00


I've posted this question many times.  I am very interested in knowing
myself.  Although, I'm not interested in saving miliseconds here or there.
However, if I can cut off significant processing time by switching from
dynamic to static, then I would.  Nobody has ever prooved this to that
extent.  The zero administration of dynamic files is worth a very great
deal to me.

If anyone out there has made the switch because of performance issues,
please share !  We would all like to know (Ive been asked this many times
by many people).

Good Luck
Anthony Dzikiewicz



Quote:> I don't want to provoke another monster thread on this. Can anyone point
me
> at a documentation/discussion regarding the performance penalties
associated
> with dynamic files in Universe.

> I don't necessarily want a discussion of linear hashing algorithms or
file
> structure, I'm looking for (ideally) numbers showing comparitive
> read/update/select times for dynamic vs static, particularly for files
> exceeding 1GB/1M records.

> Barry.


 
 
 

Dynamic Files vs Static Hashed - Performance Issues

Post by Clint Byru » Sat, 12 Jun 1999 04:00:00


This doesn't answer the question directly, but here's my two bits.


>I've posted this question many times.  I am very interested in knowing
>myself.  Although, I'm not interested in saving miliseconds here or there.
>However, if I can cut off significant processing time by switching from
>dynamic to static, then I would.  Nobody has ever prooved this to that
>extent.

>The zero administration of dynamic files is worth a very great
>deal to me.

I agree. Not having to worry about file sizing is a big issue. I consulted
for a company whose 'pick guy' had split on them just 8 months prior. Some
of the history files were already 30-40% in overflow.

 What are you wanting to perform better? The system as a whole or batch
processing? You might consider Indexing your files, That will speed up a lot
of the SELECTing that may be slowing your system down.

>If anyone out there has made the switch because of performance issues,
>please share !  We would all like to know (Ive been asked this many times
>by many people).

>Good Luck
>Anthony Dzikiewicz



>> I don't want to provoke another monster thread on this. Can anyone point
>me
>> at a documentation/discussion regarding the performance penalties
>associated
>> with dynamic files in Universe.

>> I don't necessarily want a discussion of linear hashing algorithms or
>file
>> structure, I'm looking for (ideally) numbers showing comparitive
>> read/update/select times for dynamic vs static, particularly for files
>> exceeding 1GB/1M records.

>> Barry.

 
 
 

Dynamic Files vs Static Hashed - Performance Issues

Post by Anthony W. Youngma » Sun, 13 Jun 1999 04:00:00




Quote:>I've posted this question many times.  I am very interested in knowing
>myself.  Although, I'm not interested in saving miliseconds here or there.
>However, if I can cut off significant processing time by switching from
>dynamic to static, then I would.  Nobody has ever prooved this to that
>extent.  The zero administration of dynamic files is worth a very great
>deal to me.

>If anyone out there has made the switch because of performance issues,
>please share !  We would all like to know (Ive been asked this many times
>by many people).

I'm busy writing the MaVerick file system at the moment. Everything's
dropping quite neatly into place, and while I (obviously) can't speak
for any other version, MaVerick will use 95% the same code across all
file-types. So about the only overhead it will suffer is the automatic
splitting/joining. And even then that won't impact performance much, as
it'll have an "intelligent" mechanism that most of the time will even
suppress whether it's worth checking for a split/merge (on the
assumption that it isn't).
--
Anthony W. Youngman - wol at thewolery dot demon dot co dot uk
Trousers with a single hole in their waistband are topologically equivalent
to a doughnut. These sugarcoated trousers have yet to catch on at fast-food
outlets! (SuperStrings by F. David Peat)

If replying by e-mail please mail wol. Anything else may get missed amongst
the spam.

 
 
 

Dynamic Files vs Static Hashed - Performance Issues

Post by Anthony Dzikiewic » Wed, 16 Jun 1999 04:00:00


Im really not looking to speed up anything actually.  However, if I knew that
converting from dynamic to static would make things go much faster in general.
Then, it would be worth considering.  I feel that there would be no considerable
difference though.  It would probably just make my job a little more demanding.
It would be just one more thing that I would have to baby sit on a regular
basis.

> This doesn't answer the question directly, but here's my two bits.


> >I've posted this question many times.  I am very interested in knowing
> >myself.  Although, I'm not interested in saving miliseconds here or there.
> >However, if I can cut off significant processing time by switching from
> >dynamic to static, then I would.  Nobody has ever prooved this to that
> >extent.

> >The zero administration of dynamic files is worth a very great
> >deal to me.

> I agree. Not having to worry about file sizing is a big issue. I consulted
> for a company whose 'pick guy' had split on them just 8 months prior. Some
> of the history files were already 30-40% in overflow.

>  What are you wanting to perform better? The system as a whole or batch
> processing? You might consider Indexing your files, That will speed up a lot
> of the SELECTing that may be slowing your system down.

> >If anyone out there has made the switch because of performance issues,
> >please share !  We would all like to know (Ive been asked this many times
> >by many people).

> >Good Luck
> >Anthony Dzikiewicz



> >> I don't want to provoke another monster thread on this. Can anyone point
> >me
> >> at a documentation/discussion regarding the performance penalties
> >associated
> >> with dynamic files in Universe.

> >> I don't necessarily want a discussion of linear hashing algorithms or
> >file
> >> structure, I'm looking for (ideally) numbers showing comparitive
> >> read/update/select times for dynamic vs static, particularly for files
> >> exceeding 1GB/1M records.

> >> Barry.

 
 
 

1. Dynamic vs. Static Hashed Files


[snip]

I LOVE dynamic files!  If for no other reason than they eliminate the need
for me to constantly remind clients that file maintenance is REALLY
somthing they need to do.  One of the best uses I have found for dynamic
files is with LARGE (say.. modulo > 100,000) files which would take days
to resize.  

I agree with your findings on the parameters, unless you have a file with
greatly varying record sizes, or some other strangeness, the defaults
generaly work well. These days, I always use dynamic files for all new
development (mostly under Universe) and havent had a problem with them
on any release >= 7.3

*** A PLEA TO VMARK***  Prime's RECLAIM was a wonderfull tool...hint,hint!

regards,
        Jeff

===================================================
Jeff Schasny     | DoD# 1735  '87 Yamaha FZ700    

===================================================

2. Copying data to MSSQL Server 7 using VB5

3. performance of dynamic vs. static SQL

4. ODBC Drivers for Interbase 4.0 in Win 95

5. Dynamic vs Static Stored Procedures

6. Foxpro for Windows: Double click in browse window??

7. Dynamic filter vs static filter...

8. Printer Escape Sequences

9. STATIC vs. DYNAMIC SQL and ODBC

10. Static vs Dynamic Recordsets

11. Dynamic SQL Vs Static SQL using Pro*C..

12. Dynamic vs. Static

13. Static vs. Dynamic ESQL