No, I’m not going to talk to you about snake oil or about moving to the public cloud or even a Colo. This is a very simple and VERY achievable plan to absolutely cut your data center infrastructure spend by at least 30% (or more). I’ve seen this and I’ve help enterprises do just this.

One of the main concerns (outside of moving to the cloud) I hear from the leadership of large enterprise is to optimize(cut) their current IT spend (debt). That phrase is usually a mixed bag of half baked and even less successful attempts at reducing their IT spend. I am going to talk about two straight-forward approaches every enterprise should be doing today – these are both table stakes to a cloud operating model yet the majority of enterprises are not doing them.

Sell me that snake oil –

1. Develop and market a Self-Service Portal (70% weighted savings)

2. Enable DevOps and CI/CD (30% weighted saving)

 

  1. When your IT delivers a self-service portal to business, what you are really delivering is a fully automated IaaS (usually) or PaaS solutions where a consumer can go request a VM or even a full application stack and it will no longer require fifteen bodies to click various buttons to do tasks such as setting a hostname, creating an AD account, updating an IPAM solution, creating a change ticket among the many other tasks a server buildout requires.You may be already thinking that this seems like a massive undertaking and thinking to yourself something like, “We can’t even get a phone deployed to a new employee within 2 weeks, and you want me to automate my entire build process?” Yes, that is absolutely what I want your organization to do. This isn’t a “nice to have” in 2018. This is a must in the datacenter. I fear for the success of companies that look at this as “too much” or “not worth it” and then in the next breath want to look at the public cloud and how it can help them [while keeping the same processes in place].The key [missing] piece of a self-service portal that doesn’t exist in a single enterprise I’ve worked with is the ability to break down lower environments such as Sandbox, Dev, Test, QA, UAT, etc. (among other items like a showback, full lifecycle,  etc.) This is key – The typical life cycle of a VM usually magically matches the lifecycle of the hardware that it is living on, IE, VM’s in lower environments (Non-Prod) will stick around until they are, frankly, deprecated by the infrastructure team because they haven’t been accessed in years.

    How does this save me money?

    With the ability to spin up workloads in minutes (read, minutes, not weeks or months) in your data center, all lower environments have the ability to be transient workloads – You can assign a lease, Dev – 10 days (a sprint), QA – 1 day, Pre-Prod- 5 days, etc. that will automatically reclaim the resources when the lease expires.

  2. Providing a self-service portal does more than just enable the business to click a few (less) buttons to provision workloads. The majority of Self Service Portals are built on very good APIs that can be consumed. With the ability to define entire application stacks, a CICD tool can use that application stack to spin up and spin down lower environments automatically while providing value to the application side of the house as the developer will never leave their code suite and/or source repo as all of the API calls, infrastructure spin up and down, along with unit, regression and acceptance testing is all automated. When this model has been adopted, all lower environments become transient workloads and you have successfully cut your future spend drastically.

I rank a self-service portal at 70% of the total saving as much of the heavy lifting as well as much of the optimizations will be realized in this step. When you tie in a CICD tool, the ability for the application teams to innovate and move the business to greater revenue and more success is far beyond what was even imaginable just prior to the self-service portal going live.

As an aside, #1 on average will take around 3-6 months to design and deploy. With proper marketing internally, adoption of this operating model can be fully achieved around 6-12 months after launch. (We are not going to break down old VMs just to onboard them to this model – Think Greenfield.) Integrating a CICD tool into a Self-Service Portal (#2) varies between around a month for an advanced application deployment model, to 6+ months for designing and deploying a new application deployment model.

Aside, dos, adapting this same deployment model to the public cloud usually takes about an hour worth of work initially and then is seamless as another destination to deploy you application to, while still having governance, security and visibity into your application(s) in AWS, Azure, GCP, etc.