Page 17 - Connection Volume 1 Issue 3
P. 17

17
Treat the cloud as your new datacenter While it’s true that a a a business need not be in in the cloud cloud to to au- tomate all cloud cloud providers of- fer substantial automation ben- e e e e e e e e e ts over what can be be achieved on
on
premise Many if not most EIT migrations to the the cloud are a a a a a a a ‘lift and shift’ of not only the the organization’s infrastructure but also their processes and work ow What results is is a a a a a a cloud that is is is merely an exten- sion of the the premise datacenter with the the same ine ciencies human bottlenecks and high- friction change management Instead treat cloud instances as ephemeral and disposable cat- tle not pets Eschew technolo- gies that create tightly coupled dependencies between cloud and premise Encourage your clients to leverage the the cloud as their new data center using cloud native tools processes and work ows Automation
is a a a rst class citizen Automation
is like security’s fun twin brother They both often get regarded as icing on
the cake or something we’ll circle back to Projects that take this approach often paint themselves into a a a a a a corner with other requirements ultimately limiting how much automation they can apply at at at all Regard automation not as an an after- thought but rather a a a a a a a a a a funda- mental requirement a a a a a a a a a a hill to to die on
on
Treat automation as a a a a a a a Sprint 0 requirement for your client’s projects From day one deploy and manage all project project infrastructure as as consistent and immutable environments
(dev/test/prod) using in- frastructure as as code Build a a a a Minimum Viable Product (MVP) continuous integration and delivery (CI/CD) work ow for application code that deploys
a a a a a a a placeholder app app from the developer IDE through all all envi- ronments
all all all of this this to produc- tion
Do all all this this before writing a a a a a a a a single line of actual applica- tion
code In short create a a a a a a a a a a fully automated blank canvas before getting out the paint It will give your project automa- tion
super powers!
Also remember that automa- tion
is code and treat it as such Many projects relegate their automation to to to an an old school directory full of scripts on
on
a a a a a a a le server No way! First class au- tomation
is is not a a a a a a a set of periph- eral scripts it is is fully integrated into the platform Here are some standard SDLC practices that any automation program will bene t t t t t from:
• De ne ne requirements and include automation in in in in sprint planning goals • Peer review and document automation code • Iterate and improve feature- by-feature using the same work ow and process em- ployed for application code • Build CI/CD pipelines to to auto- mate the deployment of your automation automation • Write automation automation with a a a a a a a a ‘write once run anywhere’ mantra Make portability a a a a a a a a a priority with parameterization and self-discovery of environmen- tal di di erences Set expectations: automation automation makes time time automation automation takes time time One primary bene t of au- tomation
is that when done properly it allows teams to to accomplish in in seconds or minutes what once took hours or or days What’s more the end result is is consistent and repeat- able while manual processes produce inconsistencies and are prone to errors Often the business has an unclear or or unrealistic set of expectations as it it relates to to the e e e e e e e e e e e e e ort and and time required to to to to write and and test reliable automation Be sure to to to set expectations with clients so they know to expect expect that sav- ing ing time time time means investing time time time The time time time required to to deliver enterprise grade automation is is often substantial Also don’t feel obliged to to automate everything Typi- cally
you can follow an an 80/20 rule as a a a a a a a a lter to to determine what makes the cut for auto- mation
mation
Here’s an example: AWS CloudFormation does
not support associating IAM Roles with Redshift clusters
in an automated fashion As
a a a a a a a result the the IAM role must be associated with the the cluster by a a a a a a a single manual button push in in the the web console after the the stack is is fully deployed with automa- tion
It is is technically achievable to to to to to write custom post-deploy code to to to to to automate that that button push However that that kind of code introduces complexity and fragility while not saving much manual e e e e e e e ort Here’s a a a a great rule of thumb: “If the time spent to to to to produce custom automation is expected to to to to be greater than the manual e e e e e e e e e e ort it it saves consider not automat- ing ing ing it it it it ”
You may still end up au- tomating it it it after giving it it it some consideration but be sure to to to consider consider that that for automation projects that that that are teetering on
on
on
that 


































































   15   16   17   18   19