Fw: uptime

stevan apter — 2000-06-15 01:50:00

i thought you folks might enjoy reading this message from dick dunbar
at the ibm aix research lab to the k list. dick has been experimenting
with kdb, the k database product, on some of the big iron at the lab.

kdb is an rdbms written in 100% pure k. the source is probably five
or six pages, the executable around 150k. we expect the 64 bit version
to be quite a bit smaller (around 30k). a size-limited time-limited
free version can be downloaded from the kx website.

----- Original Message -----
From: Dick Dunbar <dick_dunbar@...>
To: <kdb@...>
Sent: Wednesday, June 07, 2000 3:11 PM
Subject: Re: uptime


> If I could be permitted to expand on that a little ... the mass is
> certainly important on disks, but it is even more important in the machine.
> Like real estate, it's "location, location, location".
>
> It's not only the physics (locality) of the data, it's also about the
> locality of the code, and is the primary reason I believe that the
> competitors don't stand a chance in a shootout with kdb.
>
> Just look at the size of the executable, and it will explain a lot ...
> the whole code fits in the instruction cache line of a modern machine.
>
> By contrast, most commercial database products take an icache instruction
> fault for every 9 instructions executed. Like the old golf joke, it's:
>
> "hit the ball, and drag george".
>
> The kdb code is so tight (and I've given kdb some suggestions on making
> it even tighter on AIX) that it never faults and you get super-scalar
> instruction execution (more than one instruction per cycle) on these
> machines.
>
> It's also about the physics of thought ... the kdb code is so tight,
> that a human ... ok, a super-human ... can keep the whole scheme
> in their mind and optimize the algorithms without entering a 6 month
> development cycle. The other guys don't stand a chance to keep up
> with this.
>
> The final point is about roots; kdb has long, well established roots
> from earlier developments. How did kdb get to be so fast?
> With 20 years of expert experience on world-class problems, that's how.
>
> arthur whitney wrote:
>
> > > It's all about physics ... the more mass, the longer it takes.
> >
> > i like that.
>
> ---
> The KDB list is managed by majordomo@...
> Send a message there, containing the word 'help', for assistance, e.g. unsubscribing.
>