CLONE was extremely slooooow

[Ruslan Zasukhin]

On 3/31/12 9:58 AM, "Fabian Kneubuehl" <support at ysd.ch> wrote:

Hi Fabian,

> Hi Ruslan, Ivan
>  
> CLONE was extremely slow. I have a 4 month old DB with some segment errors
> (again)! I start the CLONE on a ProLiant server RAID5, i7 processor 4 month
> old, or in short a very potent server. The whole process runs over 20 hours!!!

One of tips can be 

For case when you do very massive operation
It can be good idea set cache quite big 50-100 MB ...

CLONE have two phases

A) copy records into new db, coverting may be ObjectPtr ilnks into FK links
B) convert links back -- and here cache can help ...


> Some words to our DB:
> This DB contains about 7¹000 patients and 50¹000 appointments, 20¹000 invoices
> and 20¹000 treatments. This 90¹000 records points all to one of the patients.
> The sizes are 1.5 GB BLB and 350 MB DAT.

    (2GB + 2GB) / (20-30 Mb/c) = near to 5 minutes


> I don¹t know what CLONE exactly does. But I think that it make some sequential
> searches or adjustments. As soon as you reach the patients you iterate for
> each of them through every of the about 100¹000 pointersŠ
>  
> If this statement is in any way correct ­ why not simply read the pointers
> into memory, make QSort and use a binary search?

> Really, Ruslan, we have the need for speed!!! It was no problem if CLONE runs
> 1 or 2 hours but 20h is ways too much. The good side, the result is correct.

* So let us know cache size

* do not trash old db

* can you give it to us ZIP, so we play with it here?

> Please, please discuss this with yourself and Ivan!

May be we can yet develop some log for clone ...
Actually just use existed one, e.g. WarningsLog ...

To drop msg after each task ...
Then we can see what tasts was slow ...


-- 
Best regards,

Ruslan Zasukhin
VP Engineering and New Technology
Paradigma Software, Inc

Valentina - Joining Worlds of Information
http://www.paradigmasoft.com

[I feel the need: the need for speed]