Re: Informix vs. Sybase vs. Oracle vs. (gasp) MS SQL Server

Pablo Sanchez wrote:
>
> In article <0v8awLAcPRVzEwpS_at_smooth1.demon.co.uk>, David Williams <djw_at_smooth1.demon.co.uk> writes:
> > >I can do a *moderate* comparison between Informix and Sybase (32 bit
> > >port):
> > >
> > >Sybase 11.0.X Informix 7.22
> > >------------- -------------
> > >o ~2 gig max for data cache o 1 gig max for data cache
> > >
> > > Drawback Informix. With a smaller data cache there's less data you
> > > can cache in memory.
> >

Primarily you would use this amount of memory for data warehouses for which Informix offers XPS, where you can have unlimited amount of memory, simply by adding more nodes.

In a recent record breaking TPC-D benchmark, Informix used 24GB of memory.

>
> I've asked this before and perhaps you can shed some light, by doing
> RA_PAGES this only configures how many I/O's to read-ahead however, I
> do not *believe, it increase the *size* of the I/O. That is, the I/O
> is still performed as 2/4K... you are just issuing more of them. Big
> problem(*)
>
> Can you track that down? Elizabeth Suto's book isn't clear on it.

According to some material I have from Informix, the engine allocates for every 100 buffers configured, a big buffer of eight pages, which will be used to read and write physically contiguous pages in a single I/O.

On a HP9000/K400 with four CPUs and data striped across several disks I have seen the overall I/O throughput measured by iostat during a sequential scan exceed 24MB/sek, which I wouldn't call a bottleneck.

-- 
Sveinn Sveinsson                      | mailto:svenni_at_strengur.is
Strengur ltd Computer Consultants     | http://www.strengur.is
Ármúla 7                              | Phone: +354-550-9000
IS-108 Reykjavik                      | Fax: +354-550-9010
Iceland