Stretch clustering with host-based mirroring
We currently see more and more "stretched clusters" with host-based mirroring solutions in the <100km range. They combine high availability with disaster tolerance thus, reducing costs.
Wouldn't host-based mirroring across different locations raise the problem of "rolling disasters" unless you use specific clustering technologies together with backend mirroring? Are there other issues that are raised?
I am, in fact, a big fan of stretch clusters
for combining traditional local HA with disaster recovery
I'm not entirely sure what you mean by rolling disasters but I will take two guesses.
You could be referring to a scenario where a disaster strikes the first site and then strikes the second. That is a risk, of course, but it's a risk no matter how far apart you place your sites. In reality, there aren't that many disasters that affect sites 100km apart. Of course, if both sites are located on the same earthquake fault, that could be a problem. Or, if both sites are located near the same seashore, a tsunami could wipe out both sites.
But both of those are just as true if the sites are hundreds of miles apart along the same seashore or fault.
The other thing you could mean by rolling disasters is the case where some data corruption gets replicated to the remote site. Just remember that replication
does NOT replace backups -- they cannot return your system to a previous point in time. It is still necessary to perform regular tape backups.
Please let me know if I've misinterpreted your question.
Evan L. Marcus
Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our
This was first published in March 2004