> My question involves bandwidth.  Is recovery from a hot site realistic?  Do
> you subscribe to both the server and client recovery?  What is your window
> of time for recovery?

My site has run one hot site test with ADSM so far. The hot site vendor
provided replacement hardware for both the ADSM server and the ADSM clients
involved in the test. All of this equipment was in a single building, with 100
million bit per second network connectivity.

The hot site contract does not cover all of the systems we would need to
recover in the event of a prolonged disaster. The vendor has warm site and
cold site space available in the same building. We are currently leaning
toward using one of these options for systems managed by my department and not
included in the hot site contract. This would enable us to arrange for 100
million bit per second connectivity between the ADSM server and the purchased
replacement hardware.

Some of the systems backed up by the ADSM server and managed by other
departments are located in the same room as the server but not part of a
coordinated disaster recovery plan. We might end up trying to restore these
over a network connection from the hot site to the other campus buildings.
This is currently a 16 million bit per second SMDS connection. Restoring some
of the larger client systems this way is an appalling prospect.

It should be fairly easy to estimate the amount of data to be restored over
each network connection and divide by the nominal bandwidth available. If this
calculation yields an intolerably long time, you clearly have a problem. If it
yields a reasonable time, you will need to start trying to find out what
fraction of the nominal bandwidth is likely to be available in practice.

Even with Ethernet connectivity, we found it worthwhile to be somewhat
selective about what we restored. When we restored Oracle database systems, we
restored only the Oracle instances used by critical business processes. We
also refrained from restoring files created by Oracle export operations.