The future of software releases is clear. Continuous delivery is here to stay. But does that mean that legacy software systems and infrastructure need to be altogether abandoned?
If you polled the people busy redefining best practices today, they'd agree that in a decade we're going to be effortlessly collaborating on highly complex systems that are updated continuously. QA will be fully automated, deployment and infrastructure tasks will be immediate, and large enterprise-wide software projects will be able to move quickly.
In the future, software releases won't feel so "manual." We won't be sitting on 50-person conference calls talking about release-related downtime and asking teams to stay up all night to deliver software. As the software industry continues to mature, creating tools designed to facilitate rapid software delivery, we'll get there, but it's going to take some time.
For Some: The Future is Now
We can see this reality already in a few widely publicized, exception cases. Companies like Flickr and Etsy laid the foundations for rapid, continuous delivery of software to production, and as larger businesses start to adopt DevOps practices we're a few years into an industry-wide rush toward more agile approaches to release management.
Teams at Twitter and Facebook are pushing to production every single day (and often more frequently than that), but the common attribute for many of the companies moving faster is that they don't have to support the legacy software systems present throughout most established businesses. While everyone agrees that continuous delivery is the goal, it remains to be seen how many organizations are going to migrate not just their newer, greenfield projects, but how they are going to accelerate existing, legacy systems.
Don't Leave Legacy Software Systems Out of the Equation
Small and medium-sized businesses in technology and media industries lead the way toward agile software delivery because they don't have to deal with the technologies you might find at a multi-national bank or a global insurance company. Legacy applications carry with them legacy databases built on technologies thatdon't lend themselves to instant deployment models, but there are tools and technologies designed to bridge these gaps.
If you run a large, revenue-generating application there's a good chance that you deal with massive databases from IBM or Oracle. It's even more likely if your system supports business at scale that your application depends on several levels of middleware and multiple databases spanning a whole range of vendors and technologies. Until you can bring these systems along for the daily release schedule there's still work to be done.
Databases Can be Agile Too
I'm picking on databases because they are always the toughest aspect of enterprise release management, and they often present the most difficult challenges to moving toward fast-paced release cadences.
When your release timelines become constrained it's always the databases that complicate everything. In a complex release you need to account for set up and tear down time when planning your testing environment's database requirements. In a production-facing software release it is always the database operations that present the most risk in the process. Databases need to be migrated in one-off processes with little room for error, and during some releases database changes almost invariably call for production downtime.
The good news is that database vendors and other companies supporting tools such as Oracle and DB2 are developing novel solutions to make databases more agile. Test environments can now use data that is continuously masked from production data, and vendors are innovating with various approaches to block storage to make it easier to track database changes and perform rollbacks instantly.
Conclusion: Don't discount legacy databases when it comes to continuous delivery. All signs are pointing toward a future that incorporates more agile data management practices that will facilitate continuous delivery models.
Follow Old Advice: Divide and Conquer
Anyone in charge of an IT budget understands that you have to choose your battles wisely. If we wanted to upgrade every system in the department at once it would be easier if we also had an infinite budget and the ability to hire resources at the snap of a finger. Needless to say, this isn't how most organizations work. In reality, our resources are constrained, and, forget about the budget, the limited resource is often people. It's tough to hire good enterprise release managers, and once you hire a good ERM it takes months (or years) for them to get enough experience to truly manage across your enterprise.
You are not going to move your entire enterprise to a continuous delivery cadence in a year. You are going to move your enterprise to a more continuous delivery cadence over a number of years, and you should create a multi-year plan to do this in phases. In the first stage select the most difficult component (often your relational databases) along with the easiest component (one of your newer web applications.) Focus your efforts on creating a model for other teams to follow once you've successfully moved to a continuous release cadence, and use these initial projects as a chance to experiment with what works and what doesn't.
Conclusion: Don't try to move everything to continuous delivery at once. Choose your battles and understand how to support mixed release models (with Plutora.)
Don't Fight Uphill: Better Yet, Don't Fight at All
When you are acting as a change-agent in the enterprise it will be tempting to talk about continuous delivery as a "revolution." This is an easy trap to fall in because technology is in an almost constant state of reinvention. At almost every large company the pattern is the same: every few years the "old" platform is being replaced by a "new" platform. In the process of selling a rewrite the organization may have brought in new management or gone through a reorg in an effort to get it right.
The secret of almost every large IT organization is that the people maintaining those legacy databases, they occupy the "high ground," and they've seen your rewrite before. The technical experts in charge of your company's critical Oracle databases have heard the arguments for moving faster several times over several years, and they are more than willing to help... but not if you keep on talking about them as being "old and antiquated."
Don't frame your movement toward continuous delivery as a "revolution" aimed at "fixing bad, legacy practices." In fact, don't use the word "legacy" at all. When you use that term you immediately tell the bulk of your IT department that what they are working on is worthless, and you'll be fighting an uphill battle against people determined to prove you wrong.
Conclusion: Don't use battle metaphors. Don't fight. Enlist the help of your existing staff and engage them in an effort to improve an existing system to make it more amenable to continuous delivery.
Future: The Right Tools, The Right Approach
This future of rapid software delivery supported by pervasive automation is a goal we'll achieve in the next decade, but getting there is going to require a lot of organization and planning. To support your transition to continuous delivery, you should use a tool like Plutora to manage projects as you move to continuous delivery while taking your legacy software systems along for the ride.