Re: [cobalt-users] Looking for Near 100% Up Time for a Very Vocal Hosting Customer

["Steve Werby" <steve-lists@xxxxxxxxxxxx>]

"Jerry Farquhar" <jerry@xxxxxxxxxxx> wrote:
> they want to find out the expense and requirements to ensure
> 100% guarantee of service 100% of the time.

You can make any guarantee you'd like, but you'll need luck on your side to
deliver.  <g>  Even 99.99% uptime is a measly 52 minutes of downtime per
year.

> Personally I think they are ridiculous and it is going to cost them a lot
> more than they bargained for.

For high availability you'll need multiple physical servers, RAID,
connections through multiple providers, etc. and you'll need mechanisms and
processes in place.  It'll be so expensive that it's not even worth
discussing further unless they have a serious budget.

> Any way I looking for feed back from anyone
> with suggestions as to what is:
>
> a.) The most practical approach
> b.) The most cost effective approach
> c.) The easiest to manage
>
>
> Here is what they are wanting:
> ==============================
>
> Scenario 1:
> -----------
> Primary web & email fails
> (it could be a result of either hardware or software problem)
>
> Desire is to have a second server pick up immediately without any
disruption
> or loss of web site content or email functionality and that employees and
> visitors would be oblivious to any failure at any time.

If the web content is static it's not that difficult as long as you sync the
primary and secondary physical servers whenever changes are made to the
primary and have a router or some other hardware that can reroute traffic to
the secondary server if necessary.  If the site is dyamic or is
database-driven it'll be more difficult.  If that's the case let us know and
I'll comment.  As far as realistic low-cost solutions to email related
problems go, you can setup additional mail servers which are lower priority
and handle incoming email if the primary mail server can't be reached.  If
there's a temporary failure the email will be queued and delivered to the
primary mail server when it's reachable.  If it's a long term failure you'd
need a mechanism/process to give the clients access to the email.

> Scenario 2:
> -----------
> Same as Scenario 1 but also includes no disruption even in the even of a
ISP
> or Internet provider outage.
>
> FYI - This part seems simple enough to accomplish by having a backup
server
> hosted at a different location on a different ISP's backbone.  The overall
> problem I see is synchronizing the servers web content and email so that
> there would be ZERO disruption or user required changes.

It's not as simple as you think.  You would need a mechanism for immediately
and automatically recognizing the problem *and* you'd need a solution to
handle DNS caching to prevent the clients from accessing the the machine on
the IP that is unavailable.  It's tricker if you have to guarantee that all
users (not just the clients) can access the site.  It's actually easier if
you have multiple machines at the same data center, each connected through a
different provider, behind a router and other hardware/software that can
route traffic appropriately.

> I'm all ears as to ideas and suggestions..

Feel free to provide more details, but if you have to guarantee very close
to 100% availability per scenarios 1 or 2 you'll be looking at thousands of
dollars for the equipment (nevermind consulting).  You may be better off
cutting the client lose or outsourcing to a vendor who specializes in high
availability hosting.  Perhaps you can coax the customer into agreeing to a
reasonable lower cost solution that addresses their primary concerns
(whatever those might be).

--
Steve Werby
President, Befriend Internet Services LLC
http://www.befriend.com/