Boston Linux & Unix (BLU) Home | Calendar | Mail Lists | List Archives | Desktop SIG | Hardware Hacking SIG
Wiki | Flickr | PicasaWeb | Video | Maps & Directions | Installfests | Keysignings
Linux Cafe | Meeting Notes | Blog | Linux Links | Bling | About BLU

BLU Discuss list archive


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Critique?



> On Wed, Feb 28, 2007 at 04:39:27PM -0500, markw at mohawksoft.com wrote:
>> > On Wed, Feb 28, 2007 at 02:41:29PM -0500, markw at mohawksoft.com wrote:
>> >>
>> >> The Myth of Five Nines
>> >> http://www.mohawksoft.org/?q=node/38
>> >
>> > Multiple geographically disparate (and network-topologically,
>> > and power-grid) sites are indeed a requirement. They are not
>> > rare, however, and not necessarily budget-busting.
>>
>> Come to think of it, maybe not, if you can lease in some managed site,
>> but
>> they aren't that cheap.
>
> It All Depends, of course.
>
> Let's suppose that your company derives revenue from online services. That
> is, it's a profit center, not a cost center. How much more revenue will
> you generate from being nominally up 24x7 versus having a five hour
> maintenance window one day a week in the pre-dawn hours? And how much
> revenue do you lose when a customer expects you to be up at an odd time
> but you aren't?

Five hours once a week is quite high, and probably qualifies as a strawman
argument. A very high number with a huge safety net is an average of 1
hour every month or two.

There is an issue with the "lost" customer argument in that a customer
unable to access a site for a brief time is not necessarily lost.

Also, "profit" centers have cost. What is the percent of gross represented
by the data center.

>
> On the other hand, there are internal uses which are cost centers. Pretend
> that you run IT for a regional chain of retailers. Inventory and sales
> figures need to be tracked, but how much is it worth to be able to have
> those in real-time versus end-of-day versus end-of-week?

This is an important business decision, but it is not one affected by this
discussion.

>
> Size matters. A 50-person company thinks differently about 3
> more hires for 24x7 ops than a 500-person company.

To a point. Unnecessary hires are always unwanted.

>
> These calculations will tell you how much it's worth spending on increased
> uptime. Some businesses need it, some won't. Some commitments to 24x7
> service will make others cheaper, since you may have already sunk the
> cost of building out a second datacenter, or a 24x7 operations crew.

You are missing the point, and this is sort of the point I tried to make.
You can't rely on a 99.999% uptime from a data center.

It is sort of like the joke, buying a lottery ticket only slightly
increases your chance of winning. My systems have had uptimes that exceed
years, and they aren't even anything special.


>
> And there are different levels of costs. If your net presence is
> three webservers and a database, even a complete doubling of
> hardware in a far off location may not be too expensive. If it
> is, perhaps a virtual host on the other coast of the US will serve
> your emergency needs by simply showing some static pages when you
> have a problem with your main site.

Very true.

>
>> Oh, no doubt, I wasn't trying to slam Microsoft, (Though it is something
>> I
>> do enjoy) I was more pointing out that even the best funded sites have
>> unexpected issues. The point is that you can't plan for every
>> eventuality.
>> Sooner or later you will have a failure of some kind.
>
> Sure, and that's why we talk about nines of reliability: to give
> an estimate of what we expect, or a measurement of how well
> we've done.

And that is sort of the fallacy of five nines. If anyone expects it, they 
are mistaken. If a site goes uninterrupted for over a year, after a
certain point, it has more to do with luck than engineering.

I would like to know if any site anywhere has ever achieved 99.999% uptime
over the course of one or two years. If no site has, then there is no
basis for any such estimate.


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.





BLU is a member of BostonUserGroups
BLU is a member of BostonUserGroups
We also thank MIT for the use of their facilities.

Valid HTML 4.01! Valid CSS!



Boston Linux & Unix / webmaster@blu.org