[AusNOG] NextDC Melbourne - Scheduled Power Maintenance -, 15th August

[Mark Smith]

+1

At least in this situation you'll likely get advanced notice your power is
going to go away. When your gear has or sufferers from an unexpected PSU or
other failure, you don't.

This fundamentally goes to the difference between what availability you
happen to get verses what availability you design and engineer to be
assured to get.

You'll eventually pay the price if you assume coincidental 100% past
availability guaratees future 100% availability. It's pretending failure
never occurs.

If you think you are or would be paying a lot for redundancy, you probably
haven't calculated the cost of down time of the service (e.g., lost
customer sales, staff sitting around being paid yet unable to do their
jobs). That cost of down time is likely to be well in excess of any
redundancy costs.

Technical infrastructure redundancy is a form of insurance. Would you go
without theft, fire and other types of business insurance just because it
has never happened to you before?



On 11 Aug. 2017 04:54, "Mark Newton" <newton at atdot.dotat.org> wrote:

As someone who has run colo facilities before:

It should be made completely clear during customer onboarding that the
facility operator can temporarily shut down a power feed at any time.
Advance warning should be desirable but optional.

If there is risk involved in doing that, it’s your job to mitigate it. The
facility operator doesn’t know which bits of equipment in your tenancy are
critical, and they’ve already told you to dual-feed where possible and use
a rack-mount ATS for single-corded equipment.

Power work should be done during business hours, because it’s almost
impossible to get emergency support or source replacement equipment out of
hours.

If it isn’t safe to take a power feed offline during business hours, then
you (the customer) have a design problem to solve.

  - mark




> On Aug 10, 2017, at 6:56 AM, Nathan Brookfield <
Nathan.Brookfield at simtronic.com.au> wrote:
>
> Chad,
>
> That's all well and good but when you're paying a premium price for
services of this fashion you expect a certain level of service.  There is a
risk no matter what when switching from power supplies taking extra load
they would not usually take as well as swing load issues with PDU's.
>
> I completely agree with your sentiment but the risk is not to be ignored
especially during those times.
>
> Kindest Regards,
> Nathan Brookfield (VK2NAB)
>
> -----Original Message-----
> From: AusNOG [mailto:ausnog-bounces at lists.ausnog.net] On Behalf Of Chad
Kelly
> Sent: Thursday, August 10, 2017 2:54 PM
> To: ausnog at lists.ausnog.net; ausnog-request at lists.ausnog.net
> Subject: Re: [AusNOG] NextDC Melbourne - Scheduled Power Maintenance -,
15th August
>
>
>
> On 8/10/2017 10:13 AM, ausnog-request at lists.ausnog.net wrote:
>> From the latest update today, It appears that most of the works are
>> being postponed for the time being
>>
>> Not all devices within racks support 2 feeds and planning is involved
>> with these devices, which is why it is crucial to receive accurate
information about any pending outages or upgrades to the NextDC DataCentres
which affect services.
> If these services are mission critical then you really should have duel
PSU units, so that when one feed gets taken offline the equipment
automatically switches to the other feed. They won't just disconnect both
feeds at once because that would be stupid and if the entire DC was offline
for too much time then that would put the owners in a rather awkward
situation legally, as after say 8 or 10 hours of downtime it wouldn't be
good for the owners lets put it that way.
> If the single PSU units are a part of a customers co-located equipment
then really your terms of service agreement should exclude liability under
your maintenance clauses.
> I don't understand why anyone would be using single PSU equipment in a DC
environment now a days when you can buy refurbished servers that come with
two PSU as standard even when you buy them without raid.
> Also for anything that is really really mission critical you should have
it hosted in multiple datacentres anyway so if something stupid does happen
that you can't control you at least still have services online as the load
balanced services would just switch.
> Regards Chad.
>
> --
> Chad Kelly
> Manager
> CPK Web Services
> Phone 03 5273 0246
> Web www.cpkws.com.au
>
> _______________________________________________
> AusNOG mailing list
> AusNOG at lists.ausnog.net
> http://lists.ausnog.net/mailman/listinfo/ausnog
> _______________________________________________
> AusNOG mailing list
> AusNOG at lists.ausnog.net
> http://lists.ausnog.net/mailman/listinfo/ausnog

_______________________________________________
AusNOG mailing list
AusNOG at lists.ausnog.net
http://lists.ausnog.net/mailman/listinfo/ausnog
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ausnog.net/pipermail/ausnog/attachments/20170812/a96eccba/attachment.html>