SEO

OTC wrote on 17.03.2007 16:25:19:
>
> Tool said:
> > If the assumption is to install on Linux, then one could simply walk
> > through
> > the boot up messages, such as what dmesg puts out, parse CPU
information,
> > or
> > also parse /proc/cpuinfo or /proc/meminfo, etc. These information
sources
> > are
> > relatively easy to find on Linux, however, it doesn't address other
> > unix-type
> > platforms, or Windows.
>
> Yes, it would be fairly easy to do with Linux, probably also easy but
> different with (each different flavour of) Unix and possibly even easy
but
> different again with Windoze.
>
> That's why, as a first approximation, I'd suggest a handful of
half-decent
> options, it's easier than trying to suss every possible platform
> accurately.
>
> > Might want a combination of a recommended choice based on your system
or
> > choose from a list.
> >
> > -t-
> >
> > Obnoxio The Clown wrote:
> >> DL Redden said:
> >>> Wouldn't it be nice if the installer was smart enough to sniff out
the
> >>> machine info and suggest a reasonable list of configurations to
choose
> >>> from based on the machine size. That would be most useful for
getting
> >>> started right out of the box.
> >>
> >> Although it's possible to write something that probed a box's config,
it
> >> would be a lot of very platform-specific maintenance. Maybe we could
> >> write
> >> something on a GPL basis as a community with designated maintenance
> >> teams
> >> for each supported platform ... ?
> >>

Looking at this (and similar issues) from a different angle ...:

As a professional at work (in my profession that is) I want control.
And many options, parameters, etc. provide a lot of control (assuming
they work . That is control for the professional who knows what he
is doing. Ideally the stuff works somehow automatically-magically with
some default settings, so it would work for those that "do not
(exactly) know what they are doing". But I would not expect that it
works optimally for my needs. And there's nothing more annoying than
some "system" (often software) that "thinks" it knows better than
myself what I want. Such behaviour really puts me off!

A database system like IDS is a very complex system. Why is there
the expectation that such systems should work optimally in all sorts
of disparate scenarios?

Nobody expects that the average passenger can fly a Boing 767 in a
safe way. There's tangible hardware, and generally wide acceptance
that "touching earth the wrong way" is a disaster.

But with highly complex software systems? People expect an
out-of-the-box installation on any chosen hardware in any chosen
infrastructure, and of course it should scale very well for 1 user
just developing something up to thousands of users running complex
queries. And, please, all this without having to configure anything.

My guess is, such expectations prevail because there is no tangible
hardware. Nothing to really see ... so it must be easy.
But I'm sure there must be a more profound way to explain this
interesting phenomen that again and again astonishes me since many
years.

Cheers,
Martin
--
Martin Fuerderer
IBM Informix Development Munich, Germany
Information Management

Sorry, but the following text is now required by German law:
IBM Deutschland GmbH
Vorsitzender des Aufsichtsrats: Hans Ulrich Maerki
Geschäftsführung: Martin Jetter (Vorsitzender), Rudolf
Bauer, Christian Diedrich, Christoph Grandpierre,
Matthias Hartmann, Andreas Kerstan
Sitz der Gesellschaft: Stuttgart
Registergericht: Amtsgericht Stuttgart, HRB 14562
WEEE-Reg.-Nr. DE 99369940

> >
> > --
> > *
> > * http://www.youtube.com/watch?v=rVS3k...elated&search=
> > *
> >
> > _______________________________________________
> > Informix-list mailing list
> > Informix-list@iiug.org
> > http://www.iiug.org/mailman/listinfo/informix-list
> >
> > --
> > This message has been scanned for viruses and
> > dangerous content by OpenProtect(http://www.openprotect.com), and is
> > believed to be clean.
> >
>
>
> --
> Bye now,
> Obnoxio
>
> "I'm astonished anyone pays real money for this crap."
> -- Cosmo
>
> --
> This message has been scanned for viruses and
> dangerous content by OpenProtect(http://www.openprotect.com), and is
> believed to be clean.
>
> _______________________________________________
> Informix-list mailing list
> Informix-list@iiug.org
> http://www.iiug.org/mailman/listinfo/informix-list

Martin Fuerderer wrote:

> Looking at this (and similar issues) from a different angle ...:
>
> As a professional at work (in my profession that is) I want control.
> And many options, parameters, etc. provide a lot of control (assuming
> they work . That is control for the professional who knows what he
> is doing. Ideally the stuff works somehow automatically-magically with
> some default settings, so it would work for those that "do not
> (exactly) know what they are doing". But I would not expect that it
> works optimally for my needs. And there's nothing more annoying than
> some "system" (often software) that "thinks" it knows better than
> myself what I want. Such behaviour really puts me off!
>
> A database system like IDS is a very complex system. Why is there
> the expectation that such systems should work optimally in all sorts
> of disparate scenarios?
>
> Nobody expects that the average passenger can fly a Boing 767 in a
> safe way. There's tangible hardware, and generally wide acceptance
> that "touching earth the wrong way" is a disaster.
>
> But with highly complex software systems? People expect an
> out-of-the-box installation on any chosen hardware in any chosen
> infrastructure, and of course it should scale very well for 1 user
> just developing something up to thousands of users running complex
> queries. And, please, all this without having to configure anything.
>
> My guess is, such expectations prevail because there is no tangible
> hardware. Nothing to really see ... so it must be easy.
> But I'm sure there must be a more profound way to explain this
> interesting phenomen that again and again astonishes me since many
> years.
>
> Cheers,
> Martin
> --
> Martin Fuerderer
> IBM Informix Development Munich, Germany
> Information Management

I tend to agree with Martin... There is a risk that one of this
"self-prepared" installations become a production environment...

I will never forget that I was called once to help a hardware partner to
run a custom benchmark prepared by the final customer... This final
customer told me that the previous benchmark (on another hardware
vendor) was run with the default settings off the onconfig.std... I
still find this hard to believe, but their eyes got bigger as soon as
the first results popped up...

Nevertheless, I understand this need. I think we should start by
something really small (just to see stores_demo database) and have maybe
two more options...

This first option should be smaller than suggested by Obnoxio...

Quote:
"
onconfig.sandbox - 1CPU, 512MB RAM, 5GB, 1 tempdbspace of 1GB
onconfig.small - 2 CPU, 2GB RAM, 20GB, 2 temp spaces of 2.5GB
onconfig.medium - 4 CPU, 4GB RAM, 100GB, 3 temp spaces of 5GB
onconfig.large - 8 CPU, 8GB RAM, 250GB, 6 temp spaces of 10GB
"

I'd say 100-200MB of RAM and a few hundreds MB for total storage for the
first option.
Besides this, maybe a few questions about number of CPUs to use, how
much RAM and how much disk space. Them arrange this in a rational manner.

But *always* let the customer know that this is wrong, and that creating
a system with IDS or any other top RDBMS requires knowledge and planning...

I imagine that someone who pays for this kind of software should
understand this...

It's more or less the same thing as defining the correct extent size...
Any value you choose is bad.

Regards,

Fernando Nunes
Portugal

http://informix-technology.blogspot.com
My email works... but I don't check it frequently...

Martin Fuerderer wrote:

> Looking at this (and similar issues) from a different angle ...:
>
> As a professional at work (in my profession that is) I want control.
> And many options, parameters, etc. provide a lot of control (assuming
> they work . That is control for the professional who knows what he
> is doing. Ideally the stuff works somehow automatically-magically with
> some default settings, so it would work for those that "do not
> (exactly) know what they are doing". But I would not expect that it
> works optimally for my needs. And there's nothing more annoying than
> some "system" (often software) that "thinks" it knows better than
> myself what I want. Such behaviour really puts me off!
>
> A database system like IDS is a very complex system. Why is there
> the expectation that such systems should work optimally in all sorts
> of disparate scenarios?
>
> Nobody expects that the average passenger can fly a Boing 767 in a
> safe way. There's tangible hardware, and generally wide acceptance
> that "touching earth the wrong way" is a disaster.
>
> But with highly complex software systems? People expect an
> out-of-the-box installation on any chosen hardware in any chosen
> infrastructure, and of course it should scale very well for 1 user
> just developing something up to thousands of users running complex
> queries. And, please, all this without having to configure anything.
>
> My guess is, such expectations prevail because there is no tangible
> hardware. Nothing to really see ... so it must be easy.
> But I'm sure there must be a more profound way to explain this
> interesting phenomen that again and again astonishes me since many
> years.
>
> Cheers,
> Martin
> --
> Martin Fuerderer
> IBM Informix Development Munich, Germany
> Information Management

I tend to agree with Martin... There is a risk that one of this
"self-prepared" installations become a production environment...

I will never forget that I was called once to help a hardware partner to
run a custom benchmark prepared by the final customer... This final
customer told me that the previous benchmark (on another hardware
vendor) was run with the default settings off the onconfig.std... I
still find this hard to believe, but their eyes got bigger as soon as
the first results popped up...

Nevertheless, I understand this need. I think we should start by
something really small (just to see stores_demo database) and have maybe
two more options...

This first option should be smaller than suggested by Obnoxio...

Quote:
"
onconfig.sandbox - 1CPU, 512MB RAM, 5GB, 1 tempdbspace of 1GB
onconfig.small - 2 CPU, 2GB RAM, 20GB, 2 temp spaces of 2.5GB
onconfig.medium - 4 CPU, 4GB RAM, 100GB, 3 temp spaces of 5GB
onconfig.large - 8 CPU, 8GB RAM, 250GB, 6 temp spaces of 10GB
"

I'd say 100-200MB of RAM and a few hundreds MB for total storage for the
first option.
Besides this, maybe a few questions about number of CPUs to use, how
much RAM and how much disk space. Them arrange this in a rational manner.

But *always* let the customer know that this is wrong, and that creating
a system with IDS or any other top RDBMS requires knowledge and planning...

I imagine that someone who pays for this kind of software should
understand this...

It's more or less the same thing as defining the correct extent size...
Any value you choose is bad.

Regards,

Fernando Nunes
Portugal

http://informix-technology.blogspot.com
My email works... but I don't check it frequently...

Martin,

You bring up several very valid points, and I commend you for taking
the time to write more than one or two sentences.

I think there can be an assumption of who is installing the software,
it's not going to be done by my grandmother, it's going to be done
by somebody who actually took the installation media, put it on a
system, and installed it, with _some_ knowledge that they will have
to first install, then adjust settings. That doesn't mean that you
have to be an experienced DBA just to install it, to wit, look at
how MySQL, Sybase, Oracle, and others make some reasonable assumptions
to bootstrap the software into a usable, runnable state.

What might be a useful compromise is to leave the standard installation
alone, and instead, write an "Informix Advisor" program that would allow
you to inspect the system _after_ the installation, and make recommendations
on settings. This might be more achievable in several aspects, which I'll
explain.

First, if IBM creates a standard installation without all kinds of selections
it allows a standard baseline installation that everyone can work from. As it
is right now, it's workable, and IDS will come up. But we all know it has to
be tuned after the installation. You might find more of a community effort
rolled into an 'advisor' program, that could have some basic core components,
and allow plugins for different platforms. It could allow a lot of community
input, and allow IBM to retain some kind of hands-off 'purity' that doesn't
taint the adjustments. In other words, IBM provides the baseline, the community
would provide the advisor. By IBM keeping a hands-off role, as they basically
are right now, it allows you the experienced or inexperienced DBA the opportunity
to tune based on a wide variety of tunings. For you, you might not need the
advisor program/script, but for a budding DBA it will get them into the groove
a lot faster, and help to make Informix run better, which translates into a
more positive customer and developer experience.

As usual I could be completely off-base. But I would love to see a real
open source tool that could work to help solve IDS problems based on solid
metrics and experience. My own budding experience with IDS was quite
intimidating, and I would have loved to have had some kind of advisor program
or tool that would have guided me in the right direction. To this day, some
of the settings are _still_ vague, and require an advanced perspective of
the engine.

-t-


Martin Fuerderer wrote:
>
> Looking at this (and similar issues) from a different angle ...:
>
> As a professional at work (in my profession that is) I want control.
> And many options, parameters, etc. provide a lot of control (assuming
> they work . That is control for the professional who knows what he
> is doing. Ideally the stuff works somehow automatically-magically with
> some default settings, so it would work for those that "do not
> (exactly) know what they are doing". But I would not expect that it
> works optimally for my needs. And there's nothing more annoying than
> some "system" (often software) that "thinks" it knows better than
> myself what I want. Such behaviour really puts me off!
>
> A database system like IDS is a very complex system. Why is there
> the expectation that such systems should work optimally in all sorts
> of disparate scenarios?
>
> Nobody expects that the average passenger can fly a Boing 767 in a
> safe way. There's tangible hardware, and generally wide acceptance
> that "touching earth the wrong way" is a disaster.
>
> But with highly complex software systems? People expect an
> out-of-the-box installation on any chosen hardware in any chosen
> infrastructure, and of course it should scale very well for 1 user
> just developing something up to thousands of users running complex
> queries. And, please, all this without having to configure anything.
>
> My guess is, such expectations prevail because there is no tangible
> hardware. Nothing to really see ... so it must be easy.
> But I'm sure there must be a more profound way to explain this
> interesting phenomen that again and again astonishes me since many
> years.
>
> Cheers,
> Martin
> --
> Martin Fuerderer
> IBM Informix Development Munich, Germany
> Information Management
>
> Sorry, but the following text is now required by German law:
> IBM Deutschland GmbH
> Vorsitzender des Aufsichtsrats: Hans Ulrich Maerki
> Geschäftsführung: Martin Jetter (Vorsitzender), Rudolf
> Bauer, Christian Diedrich, Christoph Grandpierre,
> Matthias Hartmann, Andreas Kerstan
> Sitz der Gesellschaft: Stuttgart
> Registergericht: Amtsgericht Stuttgart, HRB 14562
> WEEE-Reg.-Nr. DE 99369940
>