Re: Perfomance & large database
Date: Fri, 20 Apr 2001 19:03:18 -0700
Message-ID: <3ae0e95a$1_at_rpc1284.daytonoh.ncr.com>
<quered_at_esiee.fr> wrote in message news:3AE05841.26FF56E6_at_esiee.fr...
> Hi,
>
> I have to build a 'high' performance database konwing that the requests
> will always be of the same type : find an address given an ID. This is
> the ONLY resquest performed but on a very large database.
> Since I'm working under linux, I've thought of creating my own structure
> to store the data, such as using a full disk and accessing only a part
> of it from an address found in a hash table. Since the disk can be seen
> as a file, it should work well (?).
> The problem is that I fear this system might not be as performant as a
> standard database system (well optimised) . . .
>
> Can someone advice me about which database would be the best to answer
> always the same kind of very simple request ???
> Is the system I've thought of not too optimistic ???
>
The data set is completely static? There are never any changes, insertions,
or
deletions? Amazing.
Speed of access (for simple queries) is not sufficient reason to prefer a
standard
dbms. Indeed, a customized solution can probably be made to outperform a
dbms. However, if you need scalabilility, transaction support,
recoverability, or
access control, then a dbms is probably a better solution than having to
re-invent
all those capabilities yourself.
Received on Sat Apr 21 2001 - 04:03:18 CEST