When detractors argue the impossibility of doing DevOps at enterprise scale, IBM Matt Ellis has got the perfect case study to disprove that argument. It comes from looking in the mirror. At the recent IBM InterConnect conference, Ellis explained how Big Blue itself is showing how DevOps can rejuvenate even the largest software development operations by undertaking its own DevOps transformation.
If ever there was a stress-test of scale, it’d be at IBM, which has some 36,000 software engineers developing in 110 locations, Ellis said. They’re running code on top of 44,000 physical servers and 116,000 virtual machines. All told, they’re working on 519Â major product lines, 150 of which are SaaS offerings, many of them also supported by mobile apps.
“That is enterprise scale,” Ellis said, explaining that IBM has taken this massive workforce from classic development patterns, through an adoption of Agile and on to the delivery of “a true DevOps experience.”
It was all driven by the need to improve business outcomes. As Ellis explained, enterprises of all types are increasingly required to lock horns with small, disruptive competitors that are better able to react quickly to market demands.
“People are getting interested in DevOps is because of this rate and pace of change that’s being forced on organizations by smaller startups that are potentially causing disintermediation of classic enterprise businesses,” he said.
Ellis detailed some of the example drivers and outcomes that played a part in IBM’s own specific DevOps journey:
Backlog Clearing
Back in 2008, IBM realized it’s method of delivering software was hurting its abililty to deliver features in tune with timely market needs.
“We’d build a project road map that runs maybe 12 to 18 months and we were getting a backlog,” he said, explaining that clients and partners in the audience probably experienced that frustration first-hand.
Project Initiation Times
Something as simple as project initiation to decide on new features and functions could involve up to 30-dayy discussions. Now IBM is able to run two-day iteration of deciding what goes into the next sprint for projects, Ellis explained.
Build Time
Build time was dragging down the process. Now IBM has reduced the process of making sure what’s built stands up and is integrated properly down to under an hour.
“So on the same morning that I run the compilation on build, I can confirm that it is hung together and works as a package,” he said.
Ultimately, the DevOps and continuous integration journey has enabled IBM to shorten release cycles for its on-premises products from a 12-month cycle down to quarterly releases.
“We’ve shortened it to support a release cycle—if we needed it—where the train leaves every three months,” Ellis said.