Testing and Deployment: Quality Is King - Deploying and Managing a Cloud Infrastructure - page 298

Container Server
A container server is responsible for tracking lists of objects. Swift containers only know
which objects are in a container. An account server is similar to a container but instead
keeps track of containers rather than objects.
Swift uses self-healing mechanisms to crawl through the data and replace bad files or
objects with healthy replicas. It also makes available statistics of objects through tracking
containers.
Deploying a Service to the Cloud
Deploying a service to the cloud involves choosing the deployment target (the need),
gathering the requirements, and planning for the steps and processes. We will use an
example of deploying a cloud storage service to give an overview of the steps involved
in deploying a service to the cloud.
The need for storage cloud arises from at least four major factors:
■ First and foremost is the exponential growth of data (customer data, catalogs, activity
logs, machine generated logs, etc.).
■ Next is limited budgeting. It is not possible to have a separate data center for each
different project.
■ Legal requirements and regulatory compliance are important factors. For example:
anonymized customer data from a telecom operator in the United States cannot be
stored in a data center in Europe.
■ Heterogeneous storage environments becomes ubiquitous, such as SQL, NoSQL, flat
files, and so on.
A cloud storage service can help alleviate all these issues. Cloud storage gives clients the
ability to connect, manage, and interact with the underlying storage solutions. The biggest
advantage of having a cloud setup is to be able to use multiple storage technologies on a
single virtual platform obfuscated to the end user by using a single layer of middleware.
End users connect to the middleware, which can then process the user query, connect to the
correct underlying storage, retrieve the results, and present the results to the user in human-
readable format.
Deployment Steps
There are four major steps to achieving an effective deployment and maintenance scheme.
Each step is described in detail:
Step 1: Determine cloud storage use case and deployment model. The use case depends on
what the system will host and how will it be used. For each use case, the capacity, latency,
and cost are prioritized. For instance, for a production data cloud, low latency is a much
larger concern than cost. Similarly, for a backup data cloud setup, capacity and cost are
bigger concerns than latency.