Global Availibity

[Marc.Thach at radianz.com]

Joe,
This is what products such as Cisco Distributed Director will do for you.
In the case of DD it plays the role of a DNS server and performs basic
availability testing, only handing out addresses that are available.  Of
course you will need at least two of them so that they are resilient.
This is not a particular recommendation of DD, it's a product I know.  I'm
sure many other products would be suitable, offering differing levels of
complexity, you might like to look at www.vegan.net/lb for load-balancing
info.
rgds
Marc TXK
________________________________________________________________________
The views expressed are personal and do not necessarily reflect those of
the organisation providing the mail address from which this message was
sent



                                                                                                                    
                    Joseph K Gainey                                                                                 
                    <gainey at ecn.pur        To:     bind-users at isc.org                                               
                    due.edu>               cc:                                                                      
                    Sent by:               Subject:     Global Availibity                                           
                    bind-users-boun                                                                                 
                    ce at isc.org                                                                                      
                                                                                                                    
                                                                                                                    
                    23/08/2001                                                                                      
                    15:29                                                                                           
                                                                                                                    
                                                                                                                    




[this was posted through the comp.protocols.dns.bind newsgroup as well]


Okay, I've read all the bantering about how DNS was never intended
to load balance etc. so please don't scream at me when I ask this
question.

I've got sites that are geographically diverse, they are located in
different parts of the country.  Assuming that my different ISP's at each
location have network connections that are 100% available at both location
then I would only need to have a standard bind domain setup with
master/slaves.

Now the problem in in the real world the network connection provied by the
various ISP's are not 100% available and when they are not available and
we use the master/slave setups then some percentage of the time clients
attempting accessing out site won't be able to.  For example:


      Client------------[Client DNS]
                        |
      +-----------------+                       [Office]
      |                                         DNS(MASTER)
      |                                           |
      |
      +---------------------+---------------------+
  (t1)|                 (t1)|                 (t1)|
      SITEA(Seattle)    SITEB(New York) SITEC(Houston)
      DNS1(SLAVE)       DNS2(SLAVE)     DNS3(SLAVE)
      WWW(1)            WWW(2)          WWW(3)




Now if the the client requests the address of WWW it will get back the
list of IP addresses in some non-guaranteed order correct?  Each time the
client requests that WWW be resolved it will get a different ip address.

Now the scenario: my ISP in Houston has a total internet connectivity
failure leaving zero access to ANY machine at that location for 24hrs.
What
happens correct me if i'm wrong, is that 1/3 (33%) of the connections to
WWW
will fail.

We have entered into our contracts with clients that we will have 99.99%
uptime
,
if 1/3 (33%) of connection made in a 24hr period fail then this is not
99.99%
uptime.  The problem i've run into is that the only way for client's (and
thier
DNS servers) to not see the down site is to remove the down site from dns.
Not
a problem right, except I'ld rather not be called at 2am to remove
something
from dns i'ld rather have dns do it itself.

Any constructive ideas?

Joe