[AusNOG] Telstra Network Down

[Chad Kelly]

On 2/2/2017 6:50 PM, Mark Smith wrote:
>
>
> On 2 Feb. 2017 4:30 pm, "Chad Kelly" <chad at cpkws.com.au 
> <mailto:chad at cpkws.com.au>> wrote:
>
>     On 2/2/2017 3:19 PM, ausnog-request at lists.ausnog.net
>     <mailto:ausnog-request at lists.ausnog.net> wrote:
>
>         Of course when people say we have 2 core data centers, this
>         should imply no
>         data center is allowed to run over 50% capacity. It's
>         odd/strange that 3
>         active core data centers should sound so unorthodox, yet this
>         is the only
>         way to assure you can run your DCs at 65% and handle a DC
>         going black. Begs
>         the question why 4 active core DCs isn't standard architecture
>         for core
>         national infrastructure (which would assure high availability
>         under 75%
>         load), and 2x efficient in idle infrastructure.
>
>
>     I like your idea in theory.
>
>
> It's not theory. At one of the ISP's I've worked for we scaled out 
> BRASes this way. As you add units of capacity, the required redundancy 
> capacity required to cover a single unit failure reduces across all 
> the other units. It works when you can divide your problem up into 
> smaller sub-problems and distribute them across a pool.
>
> The argument sometimes used against it is that it is more devices to 
> manage. True, however that is tractable by using config templates, 
> automation and device management platforms ("software defined 
> networks"). The problems of managing many devices is not a new one if 
> you've spent any time managing fleets of desktop PCs.
>
>
>
>
>
>
>
>     But building data centres costs money and a significant amount of it.
>
>
>
>
> You get what you pay for. If you need high availability, you need to 
> be prepared to pay the price if it. If you can't afford the price, 
> then it is likely your availability requirements are greater than they 
> really need to be. Put a dollar cost against the consequence of a 
> failure, and you might find you really do need to pay the price of the 
> HA you want.
>
> If you can't afford to build DCs, you rent space in other people's to 
> meet your availability goals.
>
>
>
>     I remember when the Warrnambool exchange fire occurred, a
>     discussion was had around fire suppression and the lack of it in a
>     critical exchange for regional Victoria.
>
>     Begs the question did they have appropriate levels of fire
>     suppression equipment installed?
>
>     No good having multiple lots of equipment if its not being
>     protected from fire properly.
>
>
>
> A better architecture is one where a facility fire has a far smaller 
> impact.
>
> Your unit of expansion is your potential unit of failure. Larger units 
> of expansion, larger consequences of failure.
>
> Regards,
> Mark.
>
>
>
>
>     <snip>
>
    I get where you are coming from.

A group of us were discussing the true meaning of cloud in terms of web 
hosting the other day, I basically said that if the server isn't setup 
with load balancing across multiple DC's that it isn't really proper 
cloud hosting. It needs to be setup with high availability.

A lot of providers use the term cloud when its not.


-- 
Chad Kelly
Manager
CPK Web Services
web www.cpkws.com.au
phone 03 9013 4853

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ausnog.net/pipermail/ausnog/attachments/20170202/3e34bd43/attachment.html>