FYI:

Ran across this in the Embarcadero newsgroups:

http://www.aidaim.com/articles/cs-speed.php

I did not look into how good the actual test is and based on description it's pretty basic loop to do insert, update and delete. Not surprisingly their own product comes in first on every test. Older builds of both DBISAM and EDB were used also.

However it's interesting to see EDB and DBISAM in the same test and how much faster EDB is on almost every test they ran.

Raul

Raul


You should see the latest post from them on CodeGear - claims Nexus doen't run on 64bit Windows and that the fact that RecNo is disabled by default is a problem.

https://forums.embarcadero.com/thread.jspa?messageID=334539&tstart=0#334539

I'm waiting for a response to my post on which visual controls require RecNo

Roy Lambert

Few interesting points there regarding RecNo and accurate scrollbars (the only valid use so far for it) - and the performance hit you take by maintaining it.  

However, the vendor has definitely missed the point - most of us are looking for a stable, maintained (regular updates) and supported product and that at least for me is lot more important than  (unproven) raw speed.

Raul

Yes, the comparision between DBISAM and EDB caught my eye:

   DBISAM      EDB
Read   0.297      0.078
Edit   6.391      2.390

This is suggesting that EDB is 3.8 times faster at reading data than
DBISAM and over 2.5 times faster at editing.

Is this correct or are the tests just biased that way?

Chris Holland


On 25/03/2011 20:48, Raul wrote:
> FYI:
>
> Ran across this in the Embarcadero newsgroups: http://www.aidaim.com/articles/cs-speed.php
>
> I did not look into how good the actual test is and based on description it's pretty basic loop to do insert, update and delete. Not surprisingly their own product comes in first on every test. Older builds of both DBISAM and EDB were used also.
>
> However it's interesting to see EDB and DBISAM in the same test and how much faster EDB is on almost every test they ran.
>
> Raul
>

--
Chris Holland
[Team Elevate]

Raul


I agree about accurate scrollbars. Just now when reading your post I wondered what it would take to convert all my components with scrollbars to not having the scroll button just a long up and a long down button. I may investigate someday when I have time on my hands.

Roy Lambert

Chris


All I can say is that their tests are naff. Read down the thread a bit and you'll see a post from someone saying AidAim's products were at the bottom not the top of the list.

Roy Lambert

I did look at the source code and the tests are too trivial and not representative of a real world in all but trivial cases - it's all basically a "while not eof do something with current record".

However EDB does appear quite a bit is faster in this case. Tim has optimized EDB quite a bit over the last year+ and the results do confirm that at least in this case edb beats dbisam.

Raul


<
Chris Holland wrote:

Yes, the comparision between DBISAM and EDB caught my eye:

   DBISAM      EDB
Read   0.297      0.078
Edit   6.391      2.390

This is suggesting that EDB is 3.8 times faster at reading data than
DBISAM and over 2.5 times faster at editing.

Is this correct or are the tests just biased that way?

Chris Holland
>

Raul,

<< However it's interesting to see EDB and DBISAM in the same test and how
much faster EDB is on almost every test they ran. >>

They seem to be testing the performance of navigational access over the
wire, which EDB will probably do a bit faster than DBISAM.  But, network
latency makes such tests very unreliable.  All in all, their tests don't
seem to be worth very much in terms of telling someone what kind of
performance to expect from a given product.

--
Tim Young
Elevate Software
www.elevatesoft.com

Raul,

<< However EDB does appear quite a bit is faster in this case. Tim has
optimized EDB quite a bit over the last year+ and the results do confirm
that at least in this case edb beats dbisam.  >>

One of the reasons that EDB is faster than DBISAM with writes is that it
doesn't contain index statistics for maintaining a logical record number,
which is exactly as Thorsten Engler describes in that thread.  Logical
record number = good scrollbars (sometimes, there are exceptions), but
slower write performance.

--
Tim Young
Elevate Software
www.elevatesoft.com

Dear Tim,

> One of the reasons that EDB is faster than DBISAM with writes is that it
> doesn't contain index statistics for maintaining a logical record
> number, which is exactly as Thorsten Engler describes in that thread.
> Logical record number = good scrollbars (sometimes, there are
> exceptions), but slower write performance.

I guess you read my suggestion, right ?

--
Charalampos Michael - [Creation Power] - http://www.creationpower.gr