How relable does the Internet need to be? (Was: Re: Converged Network Threat)

joshua sahala jejs at sahala.org
Thu Feb 26 01:35:17 UTC 2004


On (25/02/04 16:30), Steve Gibbard wrote:
> 
> With that in mind, how much in the way of reliability problems is it
> reasonable to expect our users to accept?

probably something more than we tell them it will be down, but less than
we would (secretly) hope - most users tend to complain if it becomes 
uncomfortable to them and they think that calling might make it better.

> 
> If the Internet is a utility, or more generally infrastructure our society
> depends on, it seems there are a bunch of different systems to compare it
> to. 

don't forget such useful things as (snail) mail and trash collection -
we tend to accept more problems with mail (except around certain
holidays)...but if we want more reliability or responsiveness, we pay
extra (or choose a different carrier).  trash is forgiving only to the
point that it isn't making things uncomfortable, ie the stench isn't
overwhelming the can of air-freshener ;)
while it is true that we accept mobility over reliability on our cell
phones, we are becoming less and less forgiving of this (hence the race
to blanket the country with cell towers).  we compare cell servive to
landline service, and we accepted that it would take time to get better
coverage, but now it must work all the time, everywhere...

> 
> There must be some threshold for what people are willing to accept in 
> terms of residential power outages, that's somewhere above 2-3 hours 
> per year.

two or three hours a year would be wonderful here (southern florida),
but the grid is old and very succeptible to lightning (or cars) taking
out a transformer/relay/etc - i agree though, there is a threshold,
which in this case is 'configurable' in the sense that users can be
conditioned to accept worse and worse service.

> 
> So, it appears that among general infrastructure we depend on, there are
> probably the following reliability thresholds:
> 

mail - about twice as long (2-3 day first class taking 5-6), but
dependent upon the importance as perceived by the customer

trash - smell not overpowering, and bins not overflowing too badly,
presence of rats or cockroaches will reduce the threshold though ;)

> 
> How Internet service fits into that of course depends on how you're
> accessing the Net.

based somewhat upon what the customer thinks the reliability should be,
and what they are conditioned to accept - everyone here asks their
friends/coworkers who has the best dsl/cable/email/cell/etc service and
price.  this is also the reason that many of us run our own mail/web/etc
servers, so that we have a better idea of what to expect (if operator
error is going to render my email useless, i want it to be my error...)
this brings up another point, we like to be able to 'blame' the error on
someone/thing...if i hose my server, well then i'm an idiot...if my dsl
provider reloads their transit router, then they are the idiot...if the
driver in front of me is going too slow in rush hour and a semi pulls in
ahead...but i digress.
in the race to put more 9's on the company website we have created the
situation where there are (in some cases), unrealistic expectations.
these expectations have not yet been tempered by time or reality, partly
because we (network operators) have done a pretty good job of running
this internet thing in an almost reliable manner.  when something goes
wrong, we do our best to prevent that from happening again (for at least
the next month or two).
as to the question of how reliable do the users expect it to be, i
believe that it is a semi-individual thing:  as a user, i expect (or
should i say hope) it to be available when i need/want to use it, but 
as an operator, i can understand how/why it isn't (but i don't always 
like it ;) )  
the internet is as important as the service we run over it...the more
vital (or money-making), the higher the expectation - especially
when it is a service that we already have

my $0.02

/joshua
-- 
Fixing Unix is easier than living with NT.
	Jonathan Gilpin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://mailman.nanog.org/pipermail/nanog/attachments/20040225/c484eb1e/attachment.sig>


More information about the NANOG mailing list