website: Documenting use cases

This commit is contained in:
Armon Dadgar 2014-07-25 00:11:33 -04:00
parent 652c76500e
commit db464a0e83
2 changed files with 107 additions and 1 deletions

View File

@ -0,0 +1,106 @@
---
layout: "intro"
page_title: "Use Cases"
sidebar_current: "use-cases"
---
# Use Cases
Before understanding use cases, it's useful to know [what Terraform is](/intro/index.html).
This page lists some concrete use cases for Terraform, but the possible use cases are
much broader than what we cover. Due to it's extensible nature, providers and provisioners
can be added to further extend Terraform's ability to manipulate resources.
#### Heroku App Setup
Heroku is a popular PaaS for hosting web apps. Developers create an app, and then
attach add-ons, such as a database, or email provider. One of the best features it
the ability to elastically scale the number of dynos or workers. However, most
non-trivial applications quickly need many add-ons and external services.
Terraform can be used to codify the setup required for a Heroku application, ensuring
that all the required add-ons are available but it can go even further, configuring
DNSimple to use an update CNAME, or setting up CloudFlare to act as a CDN for the
app. Best of all, Terraform can do all of this without using a web interface in
under 30 seconds.
#### Multi-Tier Applications
A very common pattern is the N-tier architecture. The most common 2-tier architecture is
a pool of web servers that interface with a database tier. Extending this, additional
tiers get added for API servers, caching servers, routing meshes, etc. This pattern is used
because the tiers can be scaled indepedently and provide a seperation of concerns.
Terraform is an ideal tool for building and managing these infrastructures. Each tier can
be described as a collection of resources, and the depedencies between each tier is handled
automatically; Terraform will ensure the database tier is available before the web servers
are started and that the load balancers are aware of the web nodes. Each tier can then be
scaled easily using Terraform by modifying a single `count` configuration value. Because
the creation and provisioning of a resource is codified and automated, it becomes trivial
to build autoscaling systems to elastically scale with load.
#### Self-Service Clusters
At a certain organizational size, it becomes very challenging for a centralized
operations team to manage a large and growing infrastructure. Instead it becomes
more attractive to make "self-serve" infrastructure, allowing product teams to
manage their own infrastructure using tooling provided by the central operations team.
Using Terraform, the knowledge of how to build and scale a service can be codified
in a configuration. Terraform configurations can be shared within an organization
enabling customer teams to use the configuration as a black box and use Terraform as
a tool to manage their services.
#### Software Demos
Modern software is increasing networked and distributed. Although there exists
tools like [Vagrant](http://www.vagrantup.com/) to build virtualized environments
for demos, it is still very challenging to demo software on infrastructure providers
which more closely match production environments or require a large number of nodes.
Sofware writers can provide a Terraform configuration to create, provision and
bootstrap a demo on cloud providers like AWS. This allows end users to easily demo
the software on their own infrastructure, and even enables tweaking parameters like
cluster size to more rigorously test tools at any scale.
#### Disposable Environments
It is common practice to have both a production and staging or QA environment.
These environments are smaller clones of their production counterpart, but are
used to test new applications before releasing in production. As the production
environment grows larger and more complex, it becomes increasingly onerous to
maintain an up-to-date staging environment.
Using Terraform, the production environment can codified and then shared with
staging, QA or dev. These configurations can be used to rapidly spin up new
environments to test in, and then easily disposed of. Terraform can help tame
the difficulty of maintaining parallel environments, and makes it practical
to elastically create and destroy them.
#### Software Defined Networking
Software Defined Networking (SDN) is becoming increasingly prevalent in the
datacenter, as it provides more control to operators and developers and
allows the network to better support the applications running on top. Most SDN
implementations have a control layer and infrastructure layer.
Terraform can be used to codify the configuration for software defined networks.
This configuration can then be used by Terraform to to automatically setup and modify
settings by interfacing with the control layer. This allows configuration to be
versioned and changes to be automated. As an example, [AWS VPC](http://aws.amazon.com/vpc/)
is the largest deployments and commonly used SDN implementations, and [can be configured by
Terraform](/docs/providers/aws/r/vpc.html).
#### Resource Schedulers
In large scale infrastructures, static assignment of applications to machines
becomes increasingly challenging. To solve that problem, there are a number
of schedulers like Borg, Mesos, YARN, and Kubernetes. These can be used to
dynamically schedule Docker containers, Hadoop, Spark, and many other software
tools.
Terraform is not limited to physical providers like AWS. Resource schedulers
can be treated as a provider, allowing Terraform to request resources from them.
This allows Terraform to be used in layers: to setup the physical infrastructure
running the schedulers as well as onto the scheduled grid.

View File

@ -7,7 +7,7 @@
</li>
<li<%= sidebar_current("use-cases") %>>
<a href="/intro/index.html">Use Cases</a>
<a href="/intro/use-cases.html">Use Cases</a>
</li>
<li<%= sidebar_current("vs-other") %>>