With the launch of the new Desktop experience, efforts to remove the old desktop were put in place. Due to the slow ramp of the desktop experience (performance in various countries/markets was a challenge), the old desktop feed still lived around while most folks used the new desktop experience.
Eventually though, it came time to deprecate this service, so I worked a lot on analyzing traffic to make sure nothing was redirecting to the old homepage. A lot of links from external pages and other linkedin products were still directing to the old feed, so I had to setup a bunch of redirects/build new deeplinks to the new desktop/mobile experiences.
Since most of LinkedIn's consumer products were now on this new Ember.js application, there also was an initiative to deprecate the oldest repo/service at LinkedIn, which has existed ever since the company's founding. I was tasked to deprecate anything related to Feed here. the tasks involved in deprecating the old homepage and this old repo were similar. I looked at a bunch of graphs/monitoring tools, grepped a bunch of logs across a range of machines, read a bunch of code, looked around dependency graphs, filed tickets for code changes and redirects, came up with analytics to inform whether an endpoint could be entirely shut off if a redirect was impossible, etc.
In the end, deprecating the old desktop feed saved the company approximately 750K per year (if my memory serves me well) in running the 50+ boxes across 3 datacenters.