Re: {bb} Computing Wellness Factor ?

[Philip Clark <bblist@cannae.uklinux.net>]

On Sun, 2006-02-05 at 23:43, John Foley wrote:
> Philip, et al,
> 
> I've recently been thinking of using BB to create an
> overall "Computing Wellness Factor", which would be a
> single number that represents what BB takes several
> pages and some interpretation to convey. Now, I
> know this is somewhat simplistic, but I have my
> reasons....
> 
> Anyway, my reason in posting was to ask if you know
> if anyone has done (or tried to do) something like
> this in the past ? If so, I'd be interested in
> comparing my ideas on how this would work with
> theirs. If not, I'll most likely push forward and
> try a few things, since I don't think it would be
> a big deal to implement.
> 
> If anyone has thoughts on this, please let me know !
> (including maybe a better name for it !)
> 
>     Thanks,
> 
>       John
> 
> PS - after re-reading my post, I realize I may need to
> include a bit more information on what I really mean,
> so in a nutshell, here it is - today, BB provides
> an overall 3-level status of the entire system it
> is monitoring: red, yellow, or green. I propose adding
> a more granular overall indication (ie, a number from
> 0 to 100) that shows the status of the "stuff" being
> monitored, in order to give a better indication of how
> bad (or good) things are at a given time. In other
> words, instead of saying "currently BB shows our overall
> status as yellow", I would be able to say "currently BB
> shows our overall status as 82". Of course, the yellow
> or red (or green) indication would still be there, too -
> this would just be an additional piece of information.

If I understand correctly, I would doubt that this has been done
before. Even if something similar exists, I would expect it to be
specific to a particular application and metric (processor load
for a render farm, for example).

For a snappy name, I'd suggest "Happiness". This is because,
although it appears to be universal and simple, it's actually
wholly subjective and intangible.

For most metrics, a binary status makes sense. A port is open
or it's not. You have enough disk space or you don't. Ternary
status produces a reasonable benefit in useful information for
the additional complexity in coding and configuration. I don't
think that attaching numeric values and combining different
tests will result in any useful information.

Actually, there is a bigger danger here. It is highlighted in
your post when you say that the information on the BB pages
requires "interpretation". What you propose would require the
formalization and hard-coding of all that interpretation which,
with my gift for understatement, I would describe as non-trivial.
If there are any errors or omissions in generating the number,
you will only be helping those who are unable to interpret the
data themselves to make mistakes.

To illustrate, I started thinking of questions like these: How
happy is a desk-top machine with 1GB of free disk? Is one with
2GB of free disk happier? If so, by how much? Are these machines
happier than a mail server with 5GB of free disk? What if that
mail server also has a load average of 1.5? If that load average
persists for 30 minutes, should the mail server become less
happy over time?

At the risk of being overly pessimistic, I'd say that this would
be difficult to implement and almost impossible to implement
accurately. I'd also expect that having misleading results might
be less helpful than not having them.

To me, it sounds like a living nightmare :(

Cheers, Phil.


 
-- 
Isn't it interesting that the same people who laugh at science
fiction listen to weather forecasts and economists?
(Kelvin Throop III)

--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-==-=-=-=-=-=-=-=-=-=-=-=
To unsubscribe from this list, or to subscribe to the bb-digest list
send e-mail to mailto:majordomo@bb4.com with unsubscribe bb -and/or-
subscribe bb-digest in the BODY of the message.