I once wrote a parking sticker application for an East Coast university. If you had a faculty, staff, student or visitor sticker for the campus, it was processed using my green-screen application, which went online in 1983. The university used the mainframe program with minimal changes for about a decade, until a new client/server parking system was implemented.
Today, that sticker application exists on a nine-track tape reel hanging on my wall—and probably nowhere else.
Decommissioning the parking-sticker app was relatively straightforward for the data center team, though of course I imagine that it was emotionally traumatic. Data about the stickers was stored in several tables. One contained information about humans: name, address, phone number, relationship with the university. The other was about vehicles: make, year and color; license plate number; date of sticker assignment; sticker type and serial number; expiration date; date of cancellation. We handled some interesting exceptions. For example, some faculty were issued “floating” stickers that weren’t assigned to specific vehicles. That sort of thing.
Fortunately, the old system, while important for campus security (“Who does that car parked in a no-parking zone belong to?”), it wasn’t data that needed to be retained for any length of time for legal or compliance reasons. Shutting off the legacy application was as simple as, well, shutting off the legacy application.
(Related: Tips for building a dev/test cloud)
It’s not always that simple. Other software on campus in the 1980s—and much of the software that your team writes—needed to be retained, sometimes according to campus internal regulations, other times due to government or industry rules. How long do you need to keep payroll data? Transaction data for sales from your website? Bids for products and services, and the documentation that explains how the bids were solicited? Any time you get into financial service, aerospace, safety-oriented embedded systems, insurance, human resources, or medical, information must be retained for compliance, and must be produced on demand by auditors, queries from litigators during eDiscovery, regulatory investigations, even court subpoenas.
That can make it hard—very hard—to turn off an application you no longer need. Even if the software is recording no new transactions, retention policies might necessitate keeping it alive for years. Maybe even decades, depending on the type of data being retained, and on the regulatory requirements of your industry. Think about drug records from pharmaceutical companies, or component sourcing for automobile manufacturers.
Data and application retention has many enterprise implications. For example: Before you deploy an application and its data onto a cloud provider or SaaS platform, you should ask: Will that application and its data need to be retained? If so, will the provider still be around and provide access to it? If not, you need to make sure there’s a plan to bring the systems in-house (even if they are no longer needed) to archive the data outside the application in a way that conforms with regulatory requirements for retention and access, and then you can decommission the application.
A word of caution: I don’t know how long nine-track tapes last, especially if they are not well shielded. There’s a good chance that my 20-year-old tape is unreadable. Don’t rely upon unproven technology or suppliers for your own data archive, especially if the data must be retained for compliance purposes.
Is end-of-life decommissioning factored into your software development plans? Write me at email@example.com.