Clipper - GPF Indexing Big File

Author

GPF Indexing Big File

Rich Hamilton

2005-01-20, 3:55 pm

I've been having trouble indexing a dbf (no memo fields) that has a bit over
4.5 million records. (With that big of a file, I probably don't need to add
that it's been working fine for a long time., but size apparently got it.)

I use SIXCDX with this application, but I've also found it fails with DBFCDX
and SIXNSX.

Amazingly, it indexes just fine with DBFNTX and SIXNTX. However, it's
painfully slow and in the past I've had corruption problems with the NTX
index files and never had a corruption problem with CDX.

I'm linking with Blinker 6.0.

Has anybody else been running Clipper on dbfs this big or bigger? I know the
Clipper specs say a billion records, but what about the indexes?

Rich Hamilton
Using Walt Disney's business strategy in Your Business
www.MagicStrategy.com

Brad Schmidt

2005-01-20, 3:55 pm

Advantage Database Server should be able to reindex an index of that
size in less than a minute. You may want to look into it if you want to
significantly reduce issues related to slow performance, database
stability (including index corruption), and database security issues in
your application : www.AdvantageDatabase.com

In article <07FGd.33955$St.4071@fed1read03>, rich@sell-remove-better.com
says...
> I've been having trouble indexing a dbf (no memo fields) that has a bit over
> 4.5 million records. (With that big of a file, I probably don't need to add
> that it's been working fine for a long time., but size apparently got it.)
>
> I use SIXCDX with this application, but I've also found it fails with DBFCDX
> and SIXNSX.
>
> Amazingly, it indexes just fine with DBFNTX and SIXNTX. However, it's
> painfully slow and in the past I've had corruption problems with the NTX
> index files and never had a corruption problem with CDX.
>
> I'm linking with Blinker 6.0.
>
> Has anybody else been running Clipper on dbfs this big or bigger? I know the
> Clipper specs say a billion records, but what about the indexes?
>
> Rich Hamilton
> Using Walt Disney's business strategy in Your Business
> www.MagicStrategy.com

Brad

Al Acker

2005-01-20, 3:55 pm

Przemyslaw,

It's pretty hard to convince large corporate IT departments that they have
to install Linux servers just to keep an index from being trashed <g>.
Especially when they don't have to <G>.

Al
--
Al Acker, President
Acker Consulting, Inc.
mailto:al@thexbasefiles.com
http://www.ackerconsulting.com

Przemyslaw Czerpak

2005-01-20, 3:55 pm

On Tue, 18 Jan 2005 13:48:22 -0700,
Al Acker <al@thexbasefiles.com> wrote:
> Przemyslaw,
> It's pretty hard to convince large corporate IT departments that they have
> to install Linux servers just to keep an index from being trashed <g>.
> Especially when they don't have to <G>.

Do you think it will be easier to install NT or NOVELL and ADS? ;-)
But I agree that for simple single reindexing it's not worth of effort.
I would write simple application in xHarbour for it ;-)

best regards,
Przemek

Rich Hamilton

2005-01-20, 8:55 pm

I'm using Cliper 5.2e. I don't have Comix, but I may be able to try it.

I've also wondered about Blinker 7.0 (instead of 6.0), but the upgrade notes
don't suggest a solution.

"Markus Wiederstein" <markus.wiederstein@web.de> wrote in message
news:opskqpj4hnolrrg1@news.t-online.de...
> Am Sun, 16 Jan 2005 19:08:58 -0700 hat Rich Hamilton
> <rich@sell-remove-better.com> geschrieben:
>
>
> What clipper version do you use ?
>
> If it is 5.2e, then simply try comix.
>
> http://www.grafxsoft.com/comix.htm
>
> Markus
>
>
>
>
>
> --
> Using M2, Opera's revolutionary e-mail client: http://www.opera.com/m2/

Claudio Voskian

2005-01-20, 8:55 pm

Rich

"Rich Hamilton" <rich@sell-remove-better.com> escribió en el mensaje
news:CvQGd.36219$St.32427@fed1read03...
| I'm using Cliper 5.2e. I don't have Comix, but I may be able to try it.
|
| I've also wondered about Blinker 7.0 (instead of 6.0), but the upgrade
notes
| don't suggest a solution.
|
| "Markus Wiederstein" <markus.wiederstein@web.de> wrote in message
| news:opskqpj4hnolrrg1@news.t-online.de...
| > Am Sun, 16 Jan 2005 19:08:58 -0700 hat Rich Hamilton
| > <rich@sell-remove-better.com> geschrieben:
| >
| >
| > What clipper version do you use ?
| >
| > If it is 5.2e, then simply try comix.
| >
| > http://www.grafxsoft.com/comix.htm
| >
| > Markus
| >
| >
| > > I've been having trouble indexing a dbf (no memo fields) that has a
bit
| > > over
| > > 4.5 million records. (With that big of a file, I probably don't need
to
| > > add
| > > that it's been working fine for a long time., but size apparently got
| > > it.)
| > >
| > > I use SIXCDX with this application, but I've also found it fails with
| > > DBFCDX
| > > and SIXNSX.
| > >
| > > Amazingly, it indexes just fine with DBFNTX and SIXNTX. However, it's
| > > painfully slow and in the past I've had corruption problems with the
NTX
| > > index files and never had a corruption problem with CDX.
| > >
| > > I'm linking with Blinker 6.0.
| > >
| > > Has anybody else been running Clipper on dbfs this big or bigger? I
know
| > > the
| > > Clipper specs say a billion records, but what about the indexes?
| > >
| > > Rich Hamilton
| > > Using Walt Disney's business strategy in Your Business
| > > www.MagicStrategy.com
| > >
| > >
| >
| >
| >
| > --
| > Using M2, Opera's revolutionary e-mail client: http://www.opera.com/m2/
|
|

Maybe you can try with the following trick: copy the structure of that big
database (an empty dbf), then create the indexes, and then append one by one
the 4.5 million records... this should work (slowly, but will work).

As suggested by Markus, use Comix, but you will find the same problem near
in the future...

I made some tests with comix some time ago (and broplus, a "better" dbu).
I used a database of 4.7 million records, with just one char field of 58
with names in it, which I can index on _only_ if the key length is <= 27
[with left(charfield, 27)]; but if the database has just 4.0 million, I can
index on a key length of 33 max.

If I use 28 in the former, it inmediatly gives internal error 5304
(conventional memory exhausted, which is a non sense message because broplus
is a protected mode application). Any value above 28 (29 and so on),
initiates the indexing but gpfs in some place near the beggining.
If I use 34 in the later, it gives the 5304 internal error, and anything
above 34 starts indexing but gpfs.

With SIXCDX, everything fails almost the same; there are different
restrictions that may apply to the size of the key or the number of records,
it's just a matter of testing.

If you don't want to have any gpfs, use cl5.3b instead, with the native
dbfcdx/_dbfcdx rdd drivers.
Blinker does not change this odd behaviour of (the core in) the c5.2e
version.

HTH
--
Saluten
Claudio
Buenos Aires - Argentina
--
"Argentina crece gracias a que sus políticos y gobernantes dejan de robar
mientras duermen..." (adhiera al canje de deuda durante la noche!)

Al Acker

2005-01-20, 8:55 pm

> Do you think it will be easier to install NT or NOVELL and ADS? ;-)

Not sure if I understand the question.... you can install an ADS on either
NT or Novell.
If I had my choice for a typical business app.... I'd use C++ Builder and
ADS. And we have apps playing on NT, Win2000 Advanced and also Novell...
they all work well and are way solid.

Al
--
Al Acker, President
Acker Consulting, Inc.
mailto:al@thexbasefiles.com
http://www.ackerconsulting.com

"Przemyslaw Czerpak" <druzus@polbox.com> wrote in message
news:slrncuqubq.c3a.druzus@189-moc-1.acn.waw.pl...
> On Tue, 18 Jan 2005 13:48:22 -0700,
> Al Acker <al@thexbasefiles.com> wrote:
>
> Do you think it will be easier to install NT or NOVELL and ADS? ;-)
> But I agree that for simple single reindexing it's not worth of effort.
> I would write simple application in xHarbour for it ;-)
>
> best regards,
> Przemek

Al Acker

2005-01-20, 8:55 pm

Most corporate clients I work with already have networks in place.... so we
install ADS on what ever server they may have.... it's about a 10 min
install. You may or may not know that there is an Advantage Database Server
for Linux. If you had a choice, what client server backend database
backend would you use with xHarbour?

Al
--
Al Acker, President
Acker Consulting, Inc.
mailto:al@thexbasefiles.com
http://www.ackerconsulting.com

"Przemyslaw Czerpak" <druzus@polbox.com> wrote in message
news:slrncur5f7.d4c.druzus@189-moc-1.acn.waw.pl...
> On Tue, 18 Jan 2005 15:38:34 -0700,
> Al Acker <al@thexbasefiles.com> wrote:
>
> So you need as server with NT or NOVELL.
> For me is much easier to install and manege Linux the one of the
> above systems so IMHO is much less work to go to xHarbour and
> Linux then trying to use ADS what you suggested ;-)
>
> That's all.
>
> best regards,
> Przemek

Al Acker

2005-01-20, 8:55 pm

Przemyslaw,

I say....
> You're kind of comparing apples and oranges here.

You say...
>I'm afraid you are wrong.

But then you say...
> RMDBFCDX has only bit map filters functionality.

hummmm.... does that mean is't not as functional as the TDataSet ADS client
I'm using with Builder???

Al
--
Al Acker, President
Acker Consulting, Inc.
mailto:al@thexbasefiles.com
http://www.ackerconsulting.com

Rich Hamilton

2005-01-21, 3:55 am

Claudio,

Thanks! Your experience with 4.7 million records helped. Painfully, thanks
to third-party libs, I moved the application over to 5.3b and it runs fast
and well.

With 4.5 million records and two tags, the NTX indexes were built in a
little over 6 hours. The 5.3b DBFCDX indexes were built in 20 minutes!

Now I'm going to test by adding test records up to 12 million or so... to
see where it breaks again. At some point, I'll have to migrate it to
Advantage Database Server, I suspect. It's good to know that's an option.

Saluten
Rich
Phoenix, Arizona USA

"Claudio Voskian" <mixxell@hotmail.com> wrote in message
news:352603F4eivleU1@individual.net...
> Rich
>
> "Rich Hamilton" <rich@sellREMOVEbetter.com> escribió en el mensaje
> news:CvQGd.36219$St.32427@fed1read03...
> | I'm using Cliper 5.2e. I don't have Comix, but I may be able to try it.
> |
> | I've also wondered about Blinker 7.0 (instead of 6.0), but the upgrade
> notes
> | don't suggest a solution.
> |
> | "Markus Wiederstein" <markus.wiederstein@web.de> wrote in message
> | news:opskqpj4hnolrrg1@news.t-online.de...
> | > Am Sun, 16 Jan 2005 19:08:58 -0700 hat Rich Hamilton
> | > <rich@sell-remove-better.com> geschrieben:
> | >
> | >
> | > What clipper version do you use ?
> | >
> | > If it is 5.2e, then simply try comix.
> | >
> | > http://www.grafxsoft.com/comix.htm
> | >
> | > Markus
> | >
> | >
> | > > I've been having trouble indexing a dbf (no memo fields) that has a
> bit
> | > > over
> | > > 4.5 million records. (With that big of a file, I probably don't need
> to
> | > > add
> | > > that it's been working fine for a long time., but size apparently
got
> | > > it.)
> | > >
> | > > I use SIXCDX with this application, but I've also found it fails
with
> | > > DBFCDX
> | > > and SIXNSX.
> | > >
> | > > Amazingly, it indexes just fine with DBFNTX and SIXNTX. However,
it's
> | > > painfully slow and in the past I've had corruption problems with the
> NTX
> | > > index files and never had a corruption problem with CDX.
> | > >
> | > > I'm linking with Blinker 6.0.
> | > >
> | > > Has anybody else been running Clipper on dbfs this big or bigger? I
> know
> | > > the
> | > > Clipper specs say a billion records, but what about the indexes?
> | > >
> | > > Rich Hamilton
> | > > Using Walt Disney's business strategy in Your Business
> | > > www.MagicStrategy.com
> | > >
> | > >
> | > --
> | > Using M2, Opera's revolutionary e-mail client:
http://www.opera.com/m2/
> |
>
> Maybe you can try with the following trick: copy the structure of that big
> database (an empty dbf), then create the indexes, and then append one by
one
> the 4.5 million records... this should work (slowly, but will work).
>
> As suggested by Markus, use Comix, but you will find the same problem near
> in the future...
>
> I made some tests with comix some time ago (and broplus, a "better" dbu).
> I used a database of 4.7 million records, with just one char field of 58
> with names in it, which I can index on _only_ if the key length is <= 27
> [with left(charfield, 27)]; but if the database has just 4.0 million, I
can
> index on a key length of 33 max.
>
> If I use 28 in the former, it inmediatly gives internal error 5304
> (conventional memory exhausted, which is a non sense message because
broplus
> is a protected mode application). Any value above 28 (29 and so on),
> initiates the indexing but gpfs in some place near the beggining.
> If I use 34 in the later, it gives the 5304 internal error, and anything
> above 34 starts indexing but gpfs.
>
> With SIXCDX, everything fails almost the same; there are different
> restrictions that may apply to the size of the key or the number of
records,
> it's just a matter of testing.
>
> If you don't want to have any gpfs, use cl5.3b instead, with the native
> dbfcdx/_dbfcdx rdd drivers.
> Blinker does not change this odd behaviour of (the core in) the c5.2e
> version.
>
> HTH
> --
> Saluten
> Claudio
> Buenos Aires - Argentina
> --
> "Argentina crece gracias a que sus políticos y gobernantes dejan de robar
> mientras duermen..." (adhiera al canje de deuda durante la noche!)

Andre Monz

2005-01-21, 3:55 am

Hallo "Rich Hamilton" <rich@sell-remove-better.com>,
Du schriebst zu "GPF Indexing Big File " :

[Index Problem]

could you post the dbf structure and the index keys?
Or, if possible, provide a (zipped) dbf+cdx for download?

I would run a test with clipper 5.3.

André

Claudio Voskian

2005-01-21, 3:55 pm

Rich

"Rich Hamilton" <rich@sellREMOVEbetter.com> escribió en el mensaje
news:CMYHd.41532$St.28705@fed1read03...
| Claudio,
|
| Thanks! Your experience with 4.7 million records helped. Painfully, thanks
| to third-party libs, I moved the application over to 5.3b and it runs fast
| and well.
|
| With 4.5 million records and two tags, the NTX indexes were built in a
| little over 6 hours. The 5.3b DBFCDX indexes were built in 20 minutes!
|
| Now I'm going to test by adding test records up to 12 million or so... to
| see where it breaks again. At some point, I'll have to migrate it to
| Advantage Database Server, I suspect. It's good to know that's an option.
|
| Saluten
| Rich
| Phoenix, Arizona USA
|

Glad to see it worked, you are welcome.

Maybe 12 million records is not a good choice under clipper, I doubt it can
work with such amount of records... at least properly.
Maybe ADS is a solution, but it is too expensive for me or my customers.

CU!
--
Saluten
Claudio
Buenos Aires - Argentina
--
"Argentina crece gracias a que sus políticos y gobernantes dejan de robar
mientras duermen..."

Przemyslaw Czerpak

2005-01-21, 8:55 pm

On Thu, 20 Jan 2005 16:13:52 -0700,
Al Acker <al@thexbasefiles.com> wrote:
> Przemyslaw,
> I say....
> You say...
> But then you say...
> hummmm.... does that mean is't not as functional as the TDataSet ADS client
> I'm using with Builder???

No. For sure low level database driver does not have the functionality
of registering procedures on the server side. You totally missed the
therms.
[x]Harbour/CLIP/FlagShip applications compiled on Linux are
executed on the server side so there is no sense to register
part of them in the place where they already are.
I do not need such workarounds for network overhead like in ADS
because my whole application is executed on the server side.
I know that my English is horrible (sorry) but please try
to understand me. I think that I'm talking about sth what seems to
be totally new for you - I don't want to be ugly but at least your
above question suggests this.

BTW. RMDBFCDX gives AOFs functionality in ADS terminology.

best regards,
Przemek

Al Acker

2005-01-21, 8:55 pm

Przemyslaw,

Think we are having an English problem here... some times it's hard to
communicate with email ( even when both parties know English well <g>. )
When for one person English isn't their native language, it becomes even
harder <GGG>. Any way, I think we're both saying the same thing.. you say
your driver doesn't have the same functionality... and that's what I say
when I talked about apples and oranges. <g>.

Al
--
Al Acker, President
Acker Consulting, Inc.
mailto:al@thexbasefiles.com
http://www.ackerconsulting.com

"Przemyslaw Czerpak" <druzus@polbox.com> wrote in message
news:slrncv2u15.876.druzus@189-moc-1.acn.waw.pl...
> On Thu, 20 Jan 2005 16:13:52 -0700,
> Al Acker <al@thexbasefiles.com> wrote:
>
> No. For sure low level database driver does not have the functionality
> of registering procedures on the server side. You totally missed the
> therms.
> [x]Harbour/CLIP/FlagShip applications compiled on Linux are
> executed on the server side so there is no sense to register
> part of them in the place where they already are.
> I do not need such workarounds for network overhead like in ADS
> because my whole application is executed on the server side.
> I know that my English is horrible (sorry) but please try
> to understand me. I think that I'm talking about sth what seems to
> be totally new for you - I don't want to be ugly but at least your
> above question suggests this.
>
> BTW. RMDBFCDX gives AOFs functionality in ADS terminology.
>
> best regards,
> Przemek

Przemyslaw Czerpak

2005-01-21, 8:55 pm

On Fri, 21 Jan 2005 16:47:55 -0700,
Al Acker <al@thexbasefiles.com> wrote:
> Przemyslaw,
> Think we are having an English problem here... some times it's hard to
> communicate with email ( even when both parties know English well <g>. )
> When for one person English isn't their native language, it becomes even
> harder <GGG>. Any way, I think we're both saying the same thing.. you say
> your driver doesn't have the same functionality... and that's what I say
> when I talked about apples and oranges. <g>.

Yes and no :-) stand alone driver doesn't it's AOFs - whole application
does ;-).

Have nice w

end, for me it's time to go to bad ;-)

best regards,
Przemek

Przemyslaw Czerpak

2005-01-23, 3:55 am

On 18 Jan 2005 19:27:06 GMT,
Brad Schmidt <brad.schmidt.nospam@extendedsystems.com> wrote:
> Advantage Database Server should be able to reindex an index of that
> size in less than a minute. You may want to look into it if you want to
> significantly reduce issues related to slow performance, database
> stability (including index corruption), and database security issues in
> your application : www.AdvantageDatabase.com

Or move your program to Linux and [x]Harbour.
Run your program inside a screen or similar tool and you can
forget about data corruption or similar problems. Unlike in
ADS not only database but your whole application is moved to
the server side. When client station crashes your program can
still continue its job waiting for client to reconnect - he can
do that also from different computer or even place, f.e. from
home and continue the office work ;-). Slow modem connection is
enough for nice work. The application speed depends on server
performance and of course client doesn't need to use Linux - he
can work with his favorite OS.

I've just make a small test with indexing database with 1'000'000
records with random data on ATHLON@1249.382MHZ/256MB

CHARACTER FILED ("C", 20, 0)
creating index... 3.08 sec.
DATE FILED ("D", 8, 0)
creating index... 2.09 sec.
NUMERIC FILED ("N", 20, 9)
creating index... 2.45 sec.

If it's too slow for you then install as the server AMD64/3500
and 64bit [x]Harbour. The results will be 3 to 5 times better.

best regards,
Przemek

ps. The results are about 20% better then in the CVS xHarbour's
DBFCDX - I didn't commit my last modifications for indexing
large databases yet. I'll do that today or tomorrow when I'll
make some additional tests.

Przemyslaw Czerpak

2005-01-23, 3:55 am

On Tue, 18 Jan 2005 15:38:34 -0700,
Al Acker <al@thexbasefiles.com> wrote:
> Not sure if I understand the question.... you can install an ADS on either
> NT or Novell.
> If I had my choice for a typical business app.... I'd use C++ Builder and
> ADS. And we have apps playing on NT, Win2000 Advanced and also Novell...
> they all work well and are way solid.

So you need as server with NT or NOVELL.
For me is much easier to install and manege Linux the one of the
above systems so IMHO is much less work to go to xHarbour and
Linux then trying to use ADS what you suggested ;-)

That's all.

best regards,
Przemek

Rich Hamilton

2005-02-05, 3:55 pm

Updating...

Using Clipper 5.3b and the included DBFCDX, I was able to take the table up
to 15.0 million records (with randomly generated data in the fields) with
the index open. Reindexing (4 tags) took about an hour and 15 minutes. Data
access and searching (using a combination of scope and filter) were quite
acceptable, which I find quite amazing. This is all far better than using
DBFNTX, which took over 6 hours with only 4.5 million records.

Once the dbf file reached a true 2GB (2*(1024**3)), I could add data but
recno() and lastrec() failed, nailing my program. (I could still browse the
data with dclip, but the recno was ***.) I tried it with cdxlock.obj linked,
and without it linked, with same result.

(This table is a parent with key data used to look up records... the more
massive data, which I store in a child file with memo fields which I
compress before storing, is broken up into multiple files with a date-based
key that points to the appropriate file. That enables me to have an
astounding bulk of data, accessed quickly, and my client can even carry it
all around on a notebook computer which he uses in the field.)

In any case, I've learned that 5.3b is better than 5.2e for this
application, and my client has a lot of room to grow (from 4.5 million to 15
million records) before we need to fix it again.

At that time I suppose I'll need to migrate to Advantage ... and I should
probably start learning it now ... but that's assuming it can handle files
bigger than 2GB. Am I righth on that score, Al or anyone?

Rich

"Claudio Voskian" <mixxell@hotmail.com> wrote in message
news:35cbfoF4gcfi0U1@individual.net...
> Rich
>
> "Rich Hamilton" <rich@sellREMOVEbetter.com> escribió en el mensaje
> news:CMYHd.41532$St.28705@fed1read03...
> | Claudio,
> |
> | Thanks! Your experience with 4.7 million records helped. Painfully,
thanks
> | to third-party libs, I moved the application over to 5.3b and it runs
fast
> | and well.
> |
> | With 4.5 million records and two tags, the NTX indexes were built in a
> | little over 6 hours. The 5.3b DBFCDX indexes were built in 20 minutes!
> |
> | Now I'm going to test by adding test records up to 12 million or so...
to
> | see where it breaks again. At some point, I'll have to migrate it to
> | Advantage Database Server, I suspect. It's good to know that's an
option.
> |
> | Saluten
> | Rich
> | Phoenix, Arizona USA
> |
>
> Glad to see it worked, you are welcome.
>
> Maybe 12 million records is not a good choice under clipper, I doubt it
can
> work with such amount of records... at least properly.
> Maybe ADS is a solution, but it is too expensive for me or my customers.
>
> CU!
> --
> Saluten
> Claudio
> Buenos Aires - Argentina
> --
> "Argentina crece gracias a que sus políticos y gobernantes dejan de robar
> mientras duermen..."
>
>

Charles Foot

2005-02-05, 8:55 pm

If the application is 16-bit, you will still have this limitation.

Rich Hamilton wrote:
> Updating...
>
> Using Clipper 5.3b and the included DBFCDX, I was able to take the table up
> to 15.0 million records (with randomly generated data in the fields) with
> the index open. Reindexing (4 tags) took about an hour and 15 minutes. Data
> access and searching (using a combination of scope and filter) were quite
> acceptable, which I find quite amazing. This is all far better than using
> DBFNTX, which took over 6 hours with only 4.5 million records.
>
> Once the dbf file reached a true 2GB (2*(1024**3)), I could add data but
> recno() and lastrec() failed, nailing my program. (I could still browse the
> data with dclip, but the recno was ***.) I tried it with cdxlock.obj linked,
> and without it linked, with same result.
>
> (This table is a parent with key data used to look up records... the more
> massive data, which I store in a child file with memo fields which I
> compress before storing, is broken up into multiple files with a date-based
> key that points to the appropriate file. That enables me to have an
> astounding bulk of data, accessed quickly, and my client can even carry it
> all around on a notebook computer which he uses in the field.)
>
> In any case, I've learned that 5.3b is better than 5.2e for this
> application, and my client has a lot of room to grow (from 4.5 million to 15
> million records) before we need to fix it again.
>
> At that time I suppose I'll need to migrate to Advantage ... and I should
> probably start learning it now ... but that's assuming it can handle files
> bigger than 2GB. Am I righth on that score, Al or anyone?
>
> Rich
>
> "Claudio Voskian" <mixxell@hotmail.com> wrote in message
> news:35cbfoF4gcfi0U1@individual.net...
>
>
> thanks
>
>
> fast
>
>
> to
>
>
> option.
>
>
> can
>
>
>
>

Martin

2005-02-25, 3:55 pm

I find DBFCDXAX works well with files this big

Sponsored Links