Re: [gentoo-embedded] help in choosing DB engine for embedded application

[Karl Hiramoto]

On 03/30/2010 03:24 PM, Mirage ha wrote:
>
> Dear All,
>
> I do not know if this is the correct place to post this question or
> not but as you have experience in embedded field i expected
> you will help me.
>
> I facing a problem in choosing database engine for my application  my
> manager suggested to use files (e.g. txt files ) , i suggested to use
> berkeley db.
> So could you tell me which is better and if there is better solution (
> better db engine) please tell me.
> also if there is link to good database benchmark comparison please
> send it.
>
> Thank you,
> M.

Depending on the application, I might side with your manager, raw txt or 
binary files can be better in many situations.

Depends a lot on the workload, explain the application in general.

What is the most frequent operation a write, read( or search), or 
delete?   Raw files, txt or binary write() it is very fast.  Reading the 
file or seraching for something inside a large file can be slow without 
indexes, or the ability to binary search.  If the data in the file is 
already sorted, for example by  an incremeting ID, or time, binary 
search inside the file can be very fast.  Appending data to the end of a 
file can be a O(1) operation.     A DB engine after the write or delete, 
may have to update it's internal indexes, and rebalance its B-Tree, or 
update any other internal structures.


How much data? How much I/O?

Is program size important?  writing to a file is much smaller program 
footprint than a DB engine.

Optimizing for write or erase speed raw files will be faster.   
Optimizing for a fast search of data a DB engine will be faster unless 
you binary search your data in the files, or index it yourself.   If 
your going through the trouble of using your own complex indexes, then 
you might as well use a DB engine to avoid re-inventing the wheel.

--
Karl