Understanding Cloud Characteristics - Deploying and Managing a Cloud Infrastructure - page 25

Replicas
Data backup is a routine procedure within enterprises that prevents complete data loss.
Replication is a primary feature in any network-based storage system. Object-based storage
systems in the cloud are no exception. Data critical to the business would need to be repli-
cated with the latest version of the object maintained at every replicated location. The cloud
provider abstracts the replication details from the end user to enable users to choose between
low-cost storage options that have less or no replication or more-expensive object storage
offerings that have more replication.
Real-Time Replication vs. Passive Backup
Cloud tenants can choose to implement either real-time replication or passive backup,
based on the data requirements of their business. Passive backups are often the most
cost-effective options available for businesses to perform regular backups of all data on
the cloud. In a hybrid cloud deployment, the data would often be backed up on the public
cloud while being generated and consumed in the private cloud. This data archiving strat-
egy works well when data and access to it is not critical.
Data replication in real time would mean that latest version of the data object is always
available, even if one of multiple replication points face downtime. This replication is man-
aged by the cloud provider and abstracted from the tenant. Cloud tenants can enforce data
replication requirements in service-level agreements (SLAs).
Summary
A cloud is a huge pool of consolidated hardware resources that are abstracted from the OS
and the applications that are deployed on top of it. The key features of a cloud include its
ability to scale up and down based on usage and demand. Fine-grain elasticity is earmarked
by significantly lower startup costs compared to deploying your own infrastructure. The
hardware resources are completely isolated from tenants, which enables ubiquitous access
to the cloud.
With the pay-as-you-grow model, users can start up with very little cost and pay only
for the resources they utilize. A single physical server and the whole data center can
accommodate multiple tenants with fine-grain boundary/segmentation.
Tenants can spin up new instances and deploy their applications in no time with precon-
figured deployment scripts. This deployment automation enables upgrading applications
and maintenance without service disruption.