RE: High Availability -- True 7x24x365

  • From: "Pakhutkin, Maxim (Max)" <maxim.pakhutkin@xxxxxxxxxx>
  • To: "Oracle-l" <oracle-l@xxxxxxxxxxxxx>
  • Date: Thu, 7 Jul 2005 13:08:39 -0400

> And 
> "five nines" is probably way more expensive than any conceivable cost 
> that your customers are incurring due to an occasional outage of your 

That's precisely it: 

0) start with the understanding that 100% uptime is impossible (just as 100% 
security is impossible, etc)
1) evaluate what the uptime is now. Exclude scheduled downtime during off-hours 
if it doesn't cost you anything (no users on payroll waiting for the system to 
come up, scheduled jobs just pick up after the scheduled outage without a 
problem, for example).
2) evaluate how much a down time of 1% costs the enterprise over a period of 
time.
3) determine the cost of reducing the downtime by the first 1% over the same 
period of time as in 2). Hardware, support, extra personnel costs, etc. If it's 
below the cost of 2) you should try it. If not, you are fine as it is.
4) re-evaluate the above 3 periodically, at least once a year.
5) when considering buying 3rd-party systems, evaluate the uptime requirement 
for them and choose a vendor that will handle downtime most gracefully, all 
other things being equal. Design your own systems with an eye to downtime 
prevention.

The higher the cost of that 1% of downtime the more likely you will try to get 
into the 99% uptime.

Don't forget about the reverse of 1-3. If you are at 90% uptime now requiring 
$100,000 a year expense, when 10% of downtime only costs your company $10,000, 
you really want to _lower_ your uptime. Unlikely, but possible.

You can calculate down/uptime in minutes, hours, days as appropriate instead of 
percents.

It seems to me that in countries whose educational systems don't put an 
emphasis on math, cost analysis in managerial education is under-emphasized 
(comparing here my Soviet managerial education with my wife's American MBA :)).

Max Pakhutkin 

-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx]On Behalf Of Mark Bole
Sent: Thursday, July 07, 2005 11:21 AM
To: Oracle-l
Subject: Re: High Availability -- True 7x24x365


Management that has "moved into the modern digital world" would already 
understand that 100% uptime is an impossible business requirement. 
Airplanes crash, nuclear power plants trip offline, heart attack victims 
die in the emergency room -- there is no such thing as 100% uptime.  And 
"five nines" is probably way more expensive than any conceivable cost 
that your customers are incurring due to an occasional outage of your 
site.  It's frustrating when management tries to scapegoat the techies 
for its own inability or unwillingness to manage the risks of the 
business appropriately.

I was still somewhat shocked, however, when one of the largest banks in 
the world recently sent me an e-mail, as one of their credit card 
customers, informing me that their on-line account management site would 
be down for ONE FULL WEEK due to a system upgrade!!!   It's hard for me 
to imagine how even the sloppiest, most poorly run little software shop 
couldn't do better than that...

-- 
Mark Bole

Peter Barnett wrote:

> We have finally moved into the modern digital world. 
> Outages of our company web site are being noticed by
> our customers which is causing management to ask about
> maintaining 7x24x365 up time.
> 
> There are several ideas being circulated but I was
> wondering how others are doing it?
> 
> The requirement is true 7x24x365.  Patches, upgrades,
> maintenance need to be transparent to the users of our
> web sites.


--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l

Other related posts: