Today we’re introducing our new Question of the Week series, in which we'll be tapping into the expertise of our sister organization AFCOM’s Data Center Institute (DCI). Each DCI board member has been chosen for their depth of expertise, experience, and forward-thinking insights. They come from a variety of companies and organizations and work across IT, IT operations, critical facilities, data center, cloud architecture, and product/service development. This broad scope of function and experience ensures a 360-perspective on emerging data center industry trends and effects of change on each segment in the world of data center and IT infrastructure.
Want more in-depth content for data center professionals produced with the help of experts in the field? Want to connect with those experts and interact with them on a regular basis? Join AFCOM, the rapidly growing community focused on educating, connecting, and guiding the data center and IT infrastructure industry for more than 37 years.
This week, we asked DCI board members: How do you determine the right mix of infrastructure types (on-premises, colocation, public cloud, hybrid) for your workloads? What are the key factors driving those decisions? Here's what they said:
Donna Manley, IT Senior Director, University of Pennsylvania:
Looking at the magic elixir of infrastructure types is really a function of not just technical solutions but rather the "business of IT." Any one or combination of on-premises, cloud, colocation, etc. could technically work, but it’s the business elements that potentially have the greater impact:
- Business requirements: Do your due diligence in understanding the short and potentially extended use of the applications or software or service. This analysis should include areas such as business cycles, compliance, and appetite for risk.
- Security: While traditional risk tolerance may result in one type of configuration, growing cyberattack concerns present a formidable architectural challenge, regardless of platform or location.
- Sustainability: Is support for the solution going to be in-house or outsourced? If the former, take into account extra costs for things like education and tools, and whether or not you have the right resources for service success.
- Recoverability: Determine the flexibility necessary for recovery objectives, including business tolerance for service interruptions.
- Strategic direction: Does the solution being considered align with the strategic direction of the organization?
- Interoperability: Relationships that need to exist between systems or locations to allow for the transportability and integration of information or services.
- Value: Certainly there is a need to provide value, but configuration can play a large role in competitive advantage – think data.
- Cost and service pricing: Accounting 101 concepts are in play here -- cost versus benefit, ROI, Cap/Ex posture, etc. This is the culmination of all the analytics and decisions of the previous categories!
John Parker, Global Data Center Operations and Disaster Recovery, Esri:
To begin with you should create a data center strategy. Your strategy will then dictate what you do. Determining factors include: business drivers and requirements such as compliance, high availability, time-to-market, and customer requirements. IT requirements for network, security tools, support, standards, etc., along with capacity planning, future growth requirements, and budget limitations must also be addressed.
Here are a few strategy examples:
Cloud:
- Time to market
- Startups
- Unsure of size of infrastructure required
- Scale up and down as needed
Colocation
- Compliance
- High availability
- Lack of trained data center and facilities personnel
On Premises
- Sensitive data
- Efficiently managed
- Plenty of capacity
- Cheap power
Tad Davies, Senior Vice President, Bick Consulting Services:
I am not sure John's the reference to “infrastructure types” is the right choice of words. I consider the examples he stated as delivery models, to use a generic term. His company deploys a situational-based strategy for leveraging delivery models as I recall, and I believe that to be a valuable approach. One example I remember John mentioning was that new or not-yet-fully-commercialized apps are cloud-based. If deemed viable they migrate to less costly delivery-model options. If not, they are shelved.
I also think everyone needs to get away from a single-location mindset and transition to a portfolio mindset. We don’t do this with our personal investments – we use varying investment vehicles to address different issues in our lives. And we continually monitor and change as time goes on.
With the exception of startups, most organizations/companies have existing applications. An early step is to define what can and can’t be changed based upon technical and financial feasibility, as well as factoring in the risk of change.
Some questions to ask include: For a legacy application, might it be too costly to convert to cloud-capable?
- May its sunset come in the near future?
- Is it mainframe based?
- Is it simply deemed too risky to move?
- What are the factors prompting change?
- Can you assign a significance/priority to each and associate a time frame?
- Do you have a firm understanding of the TCO of each application so that you can make an informed decision?
- Does changing where the application resides affect support?
- What are security and network requirements, and are there aspects being factored into your TCO comparison?
When deciding on overall delivery models, you must also consider existing compliance issues that might dictate the model or geographic location. It’s also important to be aware early on of any corporate or board concerns or preferences.
Mike Andrea, Director, GAICD The Strategic Directions Group:
I don’t quite agree with John’s comment that you should start with the data center strategy as it will dictate what you do. The strategy should reinforce, enable, and support IT service delivery as dictated by the IT strategy (or as some now call it the cloud or hybrid cloud strategy) that in itself enables delivery and support of the strategic business plan and vision.
Unless the business has clearly articulated the what (IT, performance, apps, systems, controls, data, access), the how (owned/leased, compliance, business continuity, security, budget, governance), and the where (legal jurisdiction, cloud, local, multinational), they need systems and services delivered from an IT services perspective. The data center strategy simply becomes an infrastructure and technology document.
This chart illustrates my thinking:
Other considerations when developing a model should include:
- Understanding what you’re trying to achieve: Has the board tasked you with reducing data center, IT, applications, licensing, telecommunications, security, staff, or other costs? Are you trying to deliver a more nimble IT platform that enables new business objectives?
- Having an exit strategy: There are very few systems, platforms, or delivery models (mainframe systems being the general exception) that last more than seven years. What happens after that?
- Knowing bandwidth requirements: Can your business run in a centralised/cloud-only manner? How much bandwidth is required to enable a cloud-only solution? How big are the files being transferred each and every time a file is retrieved/saved?
- Understanding risks or impacts of telecommunications failures: What happens if you experience internet link failure – at a single site – or the whole net? Remember, the internet has experienced significant DNS and other issues over time.
- Finding out if money is an issue: What can you afford to do, or not do?
Want more in-depth content for data center professionals produced with the help of experts in the field? Want to connect with those experts and interact with them on a regular basis? Join AFCOM, the rapidly growing community focused on educating, connecting, and guiding the data center and IT infrastructure industry for more than 37 years.
