FYI Netflix is down

Jimmy Hess mysidia at gmail.com
Sat Jun 30 17:41:02 UTC 2012


On 6/30/12, Todd Underwood <toddunder at gmail.com> wrote:
> On Jun 30, 2012 11:23 AM, "Seth Mattinen" <sethm at rollernet.us> wrote:
>> But haven't they all been cascading failures?
> No.  They have not.  That's not what that term means.
>
> 'Cascading failure' has a fairly specific meaning that doesn't imply
> resilience in the face of decomposition into smaller parts.  Cascading

Not sure where you're going there;  Cascading failures are common, but
fortunately are usually temporary or have some kind of scope limit.
Cascading  just  means you have  a  dependency  between components,
where the failure of one component may result in the failure of a
second component,
the failure of the second  component results in failure of a third component,
and this process continues until no more components are dependent or no
more components are still operating.

This can happen to the small pieces inside of one specific system,
causing that system to collapse.
It's just as valid to say Cascading failure is across  across
larger/more complex pieces of different higher level systems,  where
the components of one system aren't sufficiently independent of those
in other systems,   causing both systems to fail.

Your application logic can be a point of failure,  just as readily as
your datacenter can.
Cascades can happen at a higher level where entire systems are
dependant upon entire other systems.

And it can happen Organizationally,   External dependancy risk occurs
when an entire business is dependant on another organization  (such as
product support),  to remotely administer software they sold,  and the
subcontracter of the product support org.  stops doing their job,  or
a smaller component  (one member of their staff)  becomes a
rogue/malicious element.

--
-JH




More information about the NANOG mailing list