From: "Robert Comer" 

>I would bet that Microsoft has a lot of their prestige invested in the
> "official" clustering being reliable.

That doesn't always mean much unfortunately.  I wouldn't trust them solely
over someone else's failover type clustering just because they're
Microsoft, I'd go by actual product reputation.

>While I still have a lot to
> research, that is the direction in which I am now leaning, especially
> since the mindset at my company seems to have changed to be willing to
> spend on hardware and systems... and anyway we're saving so much by
> using SQL Server for the new apps instead of Oracle (not least because
> if we went with Oracle we'd need another person just to babysit it
> fulltime), I bet we'll still come out ahead.

Agreed about Oracle, but that doesn't mean you couldn't get a perfectly
reliable product for less money. I really don't know, just saying there
might be something better and/or cheaper.  I'd do the same thing on the
AS/400 and I probably wouldn't pick IBM's offering because I know there is
something else that will work very well.

- Bob Comer



"Ellen K."  wrote in message
news:n8bu91tgbfiijuirbtcflb99ctn611vcfg{at}4ax.com...
>I would bet that Microsoft has a lot of their prestige invested in the
> "official" clustering being reliable.  While I still have a lot to
> research, that is the direction in which I am now leaning, especially
> since the mindset at my company seems to have changed to be willing to
> spend on hardware and systems... and anyway we're saving so much by
> using SQL Server for the new apps instead of Oracle (not least because
> if we went with Oracle we'd need another person just to babysit it
> fulltime), I bet we'll still come out ahead.
>
> On Wed, 1 Jun 2005 08:48:43 -0400, "Robert Comer"
>  wrote in message :
>
>>I have a feeling that the IP address switching itself isn't a great way to
>>go.  I've never had an instance where I switched out IP addresses/machines
>>and everything just worked without the users noticing anything.
>>
>>I believe there are third party failover products as well, but I don't
>>have
>>any experience with them...
>>
>>- Bob Comer
>>
>>
>>"Ellen K."  wrote in message
>>news:g9ar91ppgegcmsf59nq69v9sqc12j8jfl4{at}4ax.com...
>>> Well, then probably the "official" clustering
solution is the way to go.
>>> I haven't really delved into it yet but I did see that the IP addresses
>>> can be on different subnets.
>>>
>>> On Wed, 1 Jun 2005 04:19:33 -0400, "Robert Comer"
>>>  wrote in message
:
>>>
>>>>>I am so ignorant about this stuff.  Would that still work if the
>>>>>servers
>>>>> are in different cities?   (I suspect my boss is going
to want to have
>>>>> the redundancy between LA and TJ.)
>>>>
>>>>Routing is going to be a big problem if both servers are on different
>>>>subnets.  (a showstopper sized problem)
>>>>
>>>>- Bob Comer
>>>>
>>>>
>>>>"Ellen K." 
wrote in message
>>>>news:8kfq9198bar3hr2okk4o98dfebpdqj7qf3{at}4ax.com...
>>>>>I am so ignorant about this stuff.  Would that still work if the
>>>>>servers
>>>>> are in different cities?   (I suspect my boss is going
to want to have
>>>>> the redundancy between LA and TJ.)
>>>>>
>>>>> For that matter, if we went with clustering, would
clustering work on
>>>>> servers in different cities?
>>>>>
>>>>> On Tue, 31 May 2005 08:36:15 -0400, Mike N.

>>>>> wrote in message
:
>>>>>
>>>>>>On Mon, 30 May 2005 23:40:14 -0700, Ellen K.
>>>>>>
>>>>>>wrote:
>>>>>>
>>>>>>> It did occur to me that since we have static
routing (I think), just
>>>>>>>assigning a new IP address might not be enough.
 Is an "ARP table"
>>>>>>>where
>>>>>>>the routing instructions live?  Would new
instructions get added
>>>>>>>automatically?
>>>>>>
>>>>>>  The "ARP table" is the IP address - to
- Ethernet Hardware address
>>>>>>translation table that contains all ethernet
hardware addresses on
>>>>>>your
>>>>>>LAN
>>>>>>segment.  It is created automatically by each host
on TCP/IP as you
>>>>>>communicate with anything else.
>>>>>>
>>>>>>   One way to eliminate the user complexity of IP
switchover is to put
>>>>>> a
>>>>>>'traffic director' device in front of the 2
servers.   The device
>>>>>>would
>>>>>>detect a timeout / lack of response from the
primary and automatically
>>>>>>switch to backup.   Normally these are used for web
servers for
>>>>>>redundancy
>>>>>>to one or more web servers from a single IP.  They
may require a
>>>>>>special
>>>>>>configuration or manual failover to work for SQL
server in your
>>>>>>application.   But upon further thought - if you
are using manual
>>>>>>failover,
>>>>>>the same thing would be accomplished by a simple
router in front of
>>>>>>the
>>>>>>2
>>>>>>servers.
>>>>>
>>>>
>>>
>>
>

--- BBBS/NT v4.01 Flag-5