Practical Cloud Knowledge: Install, Configure, and Manage - Deploying and Managing a Cloud Infrastructure - page 208

Scheduling Maintenance
Maintenance is one of the best ways of preventing future problems; that is why it is better
known as preventive maintenance. However, no matter how important maintenance may
be, it often has a direct effect on the system, such as requiring total shutdown to the public
or client base or severely affecting performance because portions of the system have to be
taken down.
There really is no easy way to prevent downtime, so it's best to schedule it in such a way
that the overall effect is minimally felt by the community. The schedule would vary depend-
ing on the nature of the system and who it serves. Sometimes the weekend or nighttime is
best if the system is directly tied to a business that operates within regular office hours. If
the system is tied to leisure use or gaming, which are usually not allowed at work, then it
would be best to do maintenance during peak office hours.
But on cloud and virtualized systems, there are ways to do maintenance in the background
without the users ever noticing. Because everything exists in a virtual space, you can simply
move around the environments into other hosts with little effect to the users and then do
maintenance on the hosts from which you just migrated.
Reasons for Maintenance
It might be obvious to us why we do routing maintenance, but sometimes these reasons
are lost to us, especially if the activity becomes too routine and we forget its importance.
Reminders are often a good thing.
Performance Issues Virtual environments create a lot of metadata that the hypervisor
needs to sort out and use, and this is the same for storage servers. Programs run on guest
operating systems and sometimes encounter errors. Even RAM gets cluttered up by being
provisioned and reprovisioned to different virtual machines, not to mention all the caches
and shared storage being used. Sometimes these things just need to be flushed so that the
system can start anew, free of the clutter that once hindered it. This is the same reason we
occasionally have to clear the temporary Internet files, cookies, and caches of our browsers
or why we periodically have to restart computers that have been running for a long time.
Te s t i n g To test new features for a service, we have to use a production-like environment
invisible to the general client base. There is no problem system-testing new features in a
mostly simulated and controlled environment, but it's totally different when it's done in
a real production one. And in order for any installation of these features to take place, the
system may just have to be shut down for a short duration, which includes testing the sys-
tem on the production environment.
Upgrading Despite cloud systems being modular and the ability to add and remove stuff
without really shutting down the whole system, there are certain upgrades that really do
require everything to stop. For example, if the hypervisor is being changed to a different one
or has a major update, it might need to be shut down first, which means that the virtual envi-
ronment will be unavailable for the duration of the upgrade. Also, if the system is moving to a