At the DevOps Enterprise Summit, running in London June 5-6th, John Davis, lead architect at easyJet, will present "Calculating the Operations Cost of Software You Haven't Developed". InfoQ sat down with Davis to discuss how traditional organisations can migrate to a more collaborative "DevOps"-enabled approach for implementing IT projects, how project management and costing will change, and how microservices and automated performance testing can be leveraged to predict future costs from existing services.
Davis discussed how his DevOps Enterprise Summit (DOES) talk in London will advocate associating a project's success with customer or operational outcomes, rather than the traditional view of success being on-time and under-budget. DevOps is about finally aligning the goals of the business with IT, and the one success metric that matters is "the time between creating a business hypotheses and having the empirical data needed to either prove or disprove it".
Estimating costs for a project is difficult, whether this project runs within a traditional IT framework or a within an organisation embracing "DevOps". However, the bounded nature of microservices can make it easier to estimate costs initially, and automated performance testing can generate future costs from existing services. There must also be a no blame culture so that people are happy to raise "sunk cost" issues without recrimination, otherwise politics will prevent them for being surfaced.
Finally Davis stated that he believes there is clear value in being part of the DevOps community, and he encourages readers to attend conferences, engage in debates and share ideas. The full transcript of the interview can be found below:
InfoQ: Welcome to InfoQ, John! Could you share the executive summary for your London DevOps Enterprise Summit (DOES) talk please, and what attendees can expect to takeaway from the session?
John Davis: Sure. The summary is that estimating costs for a project is scary. This includes the operational costs and although we discuss how we can make those costs more relevant, they still need forecasting. It then goes into how using microservices can actually make it a lot easier and we discuss how through automated performance testing we can generate future costs from existing services. For people involved with large programmes these costs can be in the hundreds of thousands or even millions, so having an accurate forecast is essential.
InfoQ: You are talking about the cost for an enterprise of moving towards a 'DevOps' way of working. How do we attempt to calculate such a cost, given that the word DevOps itself is overloaded and means many things to many people?
Davis: If anything, I am talking about how a move to DevOps can give more clarity to costs and allow more traceability, from business features down to the delivery costs. On top of this I am advocating associating a project's success with customer or operational outcomes rather than coming in on-time and under-budget which is likely to result in a better use of capital.
InfoQ: How does the traditional focus on yearly budgeting/accounting factor into the challenges of costing? In your experience, do you often see the 'sunk cost fallacy' in regards to the capital locked in existing 'legacy' hardware?
Davis: Yearly budgeting is a massive factor and one I address in the talk. We need to move to a model where projects request different budgets for different phases of their lifecycle, experiment vs exploit etc. Using microservices and the approach I discuss allows teams to use empirical data to estimate the operations costs for this budget.
The "Sunk cost fallacy" does exist and can be more complicated that just misunderstanding. There needs to be a no blame culture so people are happy to raise "sunk cost" issues without recrimination, otherwise politics will prevent them for being surfaced.
InfoQ: With the rise in popularity of public cloud platforms and PaaS, what are the key data and decision points for an organistion in regards to building or buying/renting a platform?
Davis: I think that the flexibility that public cloud platforms enable has convinced most enterprises that IaaS is a no brainer. PaaS is more controversial because there is a worry of vendor lock-in. Containerisation has helped in this regard but there are still some PaaS offerings that it is hard to avoid being vendor specific such as Serverless. The number of legacy applications in the estate is an obvious influence as to the split between on-premise and cloud based solutions. Even for these, techniques such as the Strangler pattern can facilitate a migration path to the cloud.
InfoQ: What type of metrics/KPIs should an organisation be tracking before and during any migration towards a new platform (either internal or via public cloud)? What is the 'one metric that matters' that indicates an organisation has been successful?
Davis: There are two main challenges here. The first is that the "one metric that matters" will no doubt be different for each organisation. Personally I like to think of "The time between creating a business hypotheses and having the empirical data needed to either prove or disprove it" but it could be very different for individual companies.
The other is how you monitor and track some of these softer data points. However there are some very exciting things happening in the monitoring space. We are seeing more and more examples of collating heterogeneous data points together for analysis. This currently can include application, infrastructure and business. I would like to take this further to include data from the project management software. This will then allow us to produce metrics along the lines of "average time from idea to results" which would be very cool.
InfoQ: In summary, how relevant do you believe DevOps is to modern organisations looking to move faster? In your experience, what is more important in a typical enterprise space: the need for organisational change, or technological change?
Davis: I think it is incredibility relevant. Although the approach is called DevOps, I feel if anything it is finally aligning the goals of the business with IT. Maybe it should be called BusIt!
You honestly can't say that either organisational or technological change is more important than the other. If you solve one without the other, you will still fail. What is required is to have the correct structure for requirements, teams and architecture and a buy in to facilitate this at the highest levels.
InfoQ: Many thanks for your time today. Is there anything else you would like to share with the InfoQ readers?
Davis: Just to encourage them to be part of the DevOps community. Attend conferences, engage in debates and share ideas.
The DevOps Enterprise Summit London conference runs June 5-6th at the QE II Conference Centre. Additional details can be found at the IT Revolution Events website.
