Multi-Cloud Is The Default In Higher Ed
·If your enterprise is starting today, you should absolutely follow the advice of Corey Quinn and standardize on a single cloud provider. In the world of higher education, however, the enterprise began hundreds of years ago and things look a little different.
I have worked in small and large private enterprises before joining Northwestern University in 2007. I’m aware that Northwestern, a highly selective private research institution, differs in significant ways from other colleges and universities. But I’ve learned enough here to respectfully disagree about the need for a multi-cloud presence when it comes to higher education in general.
Higher Ed Is Distinct
The institution I work for was founded 171 years ago. Harvard University, which charted the course for American higher education and whose example every school is aware of and in some way responding to, traces its founding back nearly 400 years to 1636. Obviously - obviously! - these institions have grown and changed enormously in that time, and we don’t sit around talking about how John Evans and Frances Willard would have addressed today’s IT challenges. But there is an unbroken line of decisions and institutional contexts that have led us, and other schools, to where we are today.
A non-exhaustive list of the ways higher ed differs from the corporate (and even other non-profit)worlds:
- Primacy of consensus: It’s said of faculty that a vote of 99-1 is considered a tie. The committee-led decisionmaking process extends beyond the faculty senate and into the IT organization.
- Diffuse decisionmaking responsibility: In theory the division of responsibility between the board of trustees, the president, the provost, the VP for research, individual deans (and associate deans), the CIO, the executive vice president, etc. can be clearly defined and articulated. In practice these divisions and reporting lines are very hazy and relationship-based.
- Institutional mission: You wouldn’t necessarily know it from our fundraising campaigns, but the mission of a university is to create and disseminate knowledge, not to make money.
- Researcher incentives: This was the maybe the most surprising thing for me personally when I started here. Individual researchers operate more like small businesses, bringing in grant money to support their own work. They have mostly free rein to staff their labs and research teams as they see fit, and pay rent to the university in the form of “indirect costs” (a portion of the grant award money which the university keeps, the percentage varies by school) for the use of its facilities. The researchers are incentivized to bring in as much grant money as possible, and the requirements of different grants and research projects may push them towards multiple IT solutions.
- Cross-institutional collaboration: Not only will faculty members collaborate closely with their peers at other institutions, but staff will as well. I spend far more time directly sharing knowledge and collaborating with my peers at other, “rival” schools than would ever be possible in private enterprise.
- Alumni networks: A real thing. And occasionally highly placed in a vendor’s sales team.
- Students: Not only a source of brilliant new ideas and energy, but also a large source of very cheap, highly motivated, and occasionally shockingly skilled labor that turns over very quickly. (See: Alumni networks)
Multi cloud is the default
The above factors drive the vast majority of American higher education institutions to have a presence in each of the major cloud providers. For example:
The IT director of a small college within the university listens to her developers and chooses Amazon Web Services as their platform of choice to build new applications. She makes her case to the dean, who gives the thumbs up. The college is spending its own funds to run these cloud environments, and there is no policy that specifically prohibits using AWS, so away they go.
The medical school leadership decides to build a data warehouse to store clinical data from the university’s affiliated hospital so that researchers can access it safely for their projects. A new core facility (another form of small business within a university – they own equipment or resources needed by multiple faculty members and charge for its use) is created and its director hires a consulting partner to build it. The consulting partner makes a convincing case that Microsoft Azure is the right cloud to build this, and there is no policy that specifically prohibits using Azure. So away they go.
An enterprising PhD student in a molecular biology lab builds a data portal using Google Cloud to share fungus genome data, because there is no policy that specifically prohibits using Google Cloud and even if there were he’d have ignored it. The data portal quickly becomes a widely used and relied-upon resource across the entire field of fungal genomics. The student graduates but the data portal must be kept running to support the ongoing work in the field. Congratulations, the IT department now owns a fungal genomics data portal.
These are only slightly altered scenarios from my real experience and the same happens at research institutions across the country.
A valid criticism of the above scenarios is: why wasn’t the central IT cloud team in the loop to guide these decisions? And absolutely they should be, but it does require enormous time and effort to build the relationships and trust that would get the decisionmakers to include the cloud team. Because for the reasons stated above, and because of the self-service nature of cloud services in general, they are fully empowered to do it themselves.
Multi-cloud is simply a fact of life at institutions like this, and those of us in cloud governance must put in the work to make the workloads in each cloud environment as safe, secure, and cost-effective as possible.
Single cloud can happen
All that said, there are instances of similar universities that have decided to go all-in on a single cloud provider. That is only possible when there is a CIO with a clear vision, an iron will, and a time machine.
The schools that have done this made the decision very early on (think 2012-ish) in their cloud journey. I don’t want to say that moving a large research university in 2023 from a multi-cloud presence to a single cloud provider is impossible, but I would certainly want to have a very detailed conversation with anyone who was thinking of attempting this.
I would also argue that schools without a large research presence, meaning small liberal arts colleges, teaching and learning-focused public universities, etc. should, if they are thinking of using public cloud services at all, focus on a single provider for all of the reasons laid out by Corey Quinn, Lydia Leong etc.
Takeaways
A few parting thoughts:
If you work in higher ed IT and are wondering whether to try to push your institution to use a single cloud, I think it depends on where in the organization you sit. If you are in distributed/departmental IT, then yes I think it makes sense to lighten the cognitive and operational load and focus on a single cloud provider for your team or department.
If however you’re in central IT and need to keep things running smoothly for your colleagues and customers in the distributed schools and departments, you have no choice but to make the best of multicloud.
That means, at a minimum:
- Simple, clearly-defined standards that you communicate constantly and make exceptions immediately visible
- Hard preventive guardrails to prevent dangerous (and/or expensive) operations
- Regular checkins with cloud account owners
- Wearing sunscreen and getting plenty of rest