Most companies want either the rock-solid protection of a data center, or the versatile, on-demand flexibility promised by the cloud. But can you have both? There are ways to hedge your bets, of course, by adopting a hybrid IT approach – keeping a data center footprint for mission-critical workloads, and shifting other resources to the cloud. However, this approach does not really blend protection and flexibility; it simply offers protection in one place, and flexibility in another.
So, is there a medium that attempts to deliver both benefits in one location? We would argue, there is. We are seeing the emergence of a new concept many of us here at Cloud Technology Partners (CTP) call the “On-Demand Data Center.”
The term has not caught on in a big way – yet. If you Google “On-Demand Data Center,” you will come across five-year-old posts about making physical data centers more agile, and current posts about software relating to colocation facilities.
What we see is different. On a project-by-project basis, forward-looking companies are letting software development teams leverage the public cloud to stand up data center-like resources quickly, inexpensively and only when they are needed – on demand. The teams run these dedicated On-Demand Data Centers for a while, and when they decide they do not need them anymore, they delete their virtual facilities at the end of the day.
This gives companies two of the most important benefits they are looking for out of IT. Developers get the ability to embark on projects without friction. And central IT teams can take advantage of automation, providing all the governance and controls the development team needs to keep itself – and the company – safe.
Early Adopter Stage
The process is still in the early adopter stage. Many company leaders we talk to are hesitant to greenlight On-Demand Data Center initiatives because they still have lots of open questions. But a few years from now, this approach should be much more common.
Here is how On-Demand Data Centers work. After committing to one of the major public cloud vendors – AWS, Microsoft Azure or Google Cloud Platform – companies give teams the ability to create their own individual data centers in the vendor’s cloud. Using software-defined infrastructure services provided by the cloud vendor, each team stands up a highly available, redundant, fully operational data center in any region where the cloud vendor has a dedicated center.
The New Norm
At CTP we have worked with plenty of companies to build out central frameworks to operate in one or more of the major clouds. But customers use these frameworks differently than they would On-Demand Data Centers. They tend to create four walls in the public cloud, and then give departments and teams access to the overall resource. This is like renting different floors out in a ten-story building. The teams use the floors, and when they are done, the “building” is still standing.
With On-Demand Data Centers, new buildings go up every week – perhaps even every day. Teams build their buildings as big as they need, expand them, contract them and then abandon them if they choose to. A building may be ten stories today, 100 stories tomorrow and then gone by the end of the week — all managed and controlled through software-defined infrastructure.
Having this ability to dial resources up and down delivers the agility companies demand from the cloud. If, say, an insurance company wants to experiment with a new service, a team can stand up a dedicated On-Demand Data Center to house that project. If the service does not pay off quickly enough, the team can scrap that data center, and a new team can start a new project in a whole new resource.
Companies can also use On-Demand Data Centers to do a better job managing projects with contractors. Outsourcers working on software projects for banks often rent physical data centers for the duration of their engagements. This can get expensive. Being able to create an On-Demand Data Center in the cloud and close it at any time, gives the bank more control over the cost and direction of the software project. Plus, the bank can watch the behavior of the contractor and determine whether they are using the proper controls and adhering to proper design practices while developing the software.
With On-Demand Data Centers, companies can also experiment more, because the cost of failure is lower. They can commission two to three consultants to do prototypes in separate virtual data centers, and then keep the one that does the best job. Rather than pay $50,000 to $100,000 to buy data center hardware for a project, they can spend $1,000 to see if a project gets any traction by the end of the week, and, if not, bring in a new team on a new project on Monday.
Delivery of Core Services
To create an On-Demand Data Center, a company needs to enable the following set of core services:
- Account structures – Create structures that both provide barriers and link accounts together in a way that matches the organization’s business model.
- Logging – Log data needs to be moved into a repository for analysis, to give visibility into what is going on in the data center.
- Monitoring and alerting – A toolset that notifies you if there are problems, and evaluates the need for subsequent actions.
- Image pipelines – The patching that is necessary for launching services connected to software operating in the data center.
- Infrastructure pipelines – The infrastructure code used to provision the network, servers, storage and other infrastructure services needed to create a redundant and highly available platform.
- Auto-deploy pipelines – These are used by development teams to support their CICD practice. It will involve SDLC toolchains and services the team is comfortable with deploying. Also, IT can deploy multiple SDLC toolchain combinations using the same framework and On-Demand Data Center options.
- Identity and access management – Extend role-based identities out to the public cloud and the On-Demand Data Center.
- Key management system – Implement an encryption policy and manage it through tools and services.
- Security monitoring – Optics and automation to maintain consistent security posture while teams are developing and working. Look for threats and have the ability to shut the center down.
- Cost controls – Develop a governance process that prevents people from leaving resources running that are not used.
- Networking – Software-driven networking that creates firewall rules between the applications and the data, and about how people access components
- Tagging – A common tagging structure is a glue that holds together the data center and gives visibility into who is doing what
The Next Steps to Freedom
What has to happen to make On-Demand Data Centers a more common practice? Education. Teams have to see the benefits of being able to dial up resources and close them down on a dynamic basis. IT leaders need to shift their mindsets to accept the notion that data centers – just like individual servers – can be ephemeral resources.
It may take a while. But in a few years, the economic and business benefits will become clear: On-Demand Data Centers will be in demand.