4 ms·
Keep the Monolith, but Split the Workloads
- rektide 4y agoI have tried to get this done at so many places. It's a huge huge win, having monolith codedev advantages & microservices ops advantages. But I can never get believers. No one sees how obvious & great this is. We've had services where like three or four routes have absurdly terrible computation times, where they create huge backlogs of requests, for requests that could & should be low latency. Moving the slow stuff out into a different serving pool would be such a win. That's just a start on the segmentation I want, but it's such a clear calling with such huge wins. Just hearing a single other person doing this is relief. This is, to me, taking a monolith & applying the Bulkhead Pattern via infra, to make different failure/reliability/performance zones across different routes. Bulkhead pattern rules.