|
Although Octopus is a good data replication application (with exception
of open files if previous note is true), one biggie from my perspective
is that all writes (updates) now get propagated over the network. This
additional network traffic may or may not be a concern from a
performance point of view, but it also raises the issue of security. I
believe Octopus has encryption option, but this additional overhead may
be (or may not be) an issue at your site.
The other biggie is that of data consistency ie. if a primary server is
updated, but before the secondary write completes, the network link
gets dropped. The primary now has updated data that the secondary does
not know about, but the application on the primary considers the write
complete. The secondary server after not being able to reach the
primary now takes control and starts the switchover to become primary.
Boom - data inconsistency. This problem is known as the two phase
commit issue in the db world and has been extremely tough to resolve
due to the many situations where data consistency is absolutely
paramount - imagine ATM machine applying an update to one DB, but not a
mirror server in another location ..
Bottom line is that in DB world, a write is not considered complete
until ALL servers report that the update has been completed
successfully on all db's. If one server does not report success, then
ALL updates are rolled back on all servers and an error reported to the
application.
I may be wrong on this, but my understanding is that Octopus does not
support the feature of update rolbacks. This would disqualify it from
any application that required high data consistency.
[I have had some interesting conversations on Compuserve with Octopus
reps on these issues and after a few questions like this, the threads
activity goes away :-)]
Of course, for file/print application type stuff, this may not be a big
issue, but Customers should understand these type of issues before
jumping on the Octopus bandwagon.
IMO, if a Cust absolutely needs highest availability and consistency,
then they should be looking at OpenVMS cluster which allows volumes to
be shadowed across different datacenters up to 500 miles apart using
T3 /ATM lines. Volume shadowing does not consider write complete until
all volumes report update was successful.
My $.02 anyway ..
:-)
/ Kerry
Regards,
/ Kerry
Regards,
is that Octopus does not support
|