Page 37 - Building Digital Libraries

CHAPTER 2

When deciding whether cloud-based services are appropriate for your
application, consider:

which institutional resources need to connect to the service and
which policies apply. Many institutions do not allow internal
systems to connect to outside services. This could affect au-
then tication and directory services, billing, file transfer, and
integration with other locally provided services.
data regulations at your institution. Even if data is encrypted and
the vendor has no access to it, you may be prohibited from
transporting certain types of data or at least required to go
through an exception process.
size of files. It is not practical to transfer large datasets, high-
definition video, and other large resources over a network
connection outside of your institution—especially if staff or
users need to be able to manipulate these objects.

Because the term “cloud” doesn’t carry any specific meaning aside from
service being delivered over a network, it is important to base decisions on
how the service meets your needs rather than on whether or not a service
is described as being a “cloud” service.

Real or Virtual Hardware?
Recovering files is easy, but recovering systems is difficult. Unless you need
specific real hardware connected to your service, virtual hardware is usually
a better choice. Virtual environments offer the following advantages over
real hardware:

Avoid the procurement and de-accessioning processes that are
associated with physical hardware. Bids, purchase orders, ship-
ping, inventory, installation, and surplusing at end of life are
not necessary.
Greatly simplified disaster recovery. Identical hardware is not
necessary to recover from hardware failures, and entire pre-
configured systems can be launched within minutes. Testing
disaster recovery is also simplified—it is only necessary to
verify that images launch properly. Backups can be made with
services running.
Easily relocate services to other environments. Services can be
moved across departments as well as into or out of the institution.

Improve uptime.
Simpler upgrades. Successful experiments can be incorporated
into the production image, while unsuccessful ones are quickly
rolled back. Multiple different images can be kept.

22