Splitting the monolith: From Django to Falcon

on YouTube ↗16m

" summary

CMS stays in Django for authoring while a new Falcon service serves reads, backed by Falcon + Elasticsearch + Varnish and an index refresh every 30–40 minutes to avoid expensive syncs. Result: some endpoints become ~10–20x faster, peak caches handle ~8 million requests from Varnish, data set ~200,000 movies with millions of related rows. First service migration takes ~2–2.5 months; cache invalidation, replay testing and incremental deployment strategies prove effective.

AI-written summary · about

" chapters (6)

1. 0:00Start
2. 1:05Do not start micro
3. 4:04By the numbers
4. 6:51Migration obstacles
5. 10:57Tooling can go a long way
6. 11:50Advantages of Microservices

" related talks

• 45mScaling: from 0 to 20 million usersJosip Stuhli, Cal Evans · Dutch PHP Conference 2022
• 47mFrom Legacy to High-Load: Evolution of a Web ApplicationYevgen Lysenko · Highload fwdays'18
• 39mHigh Performance Django: From Runserver to Reddit HugsPeter Baumgartner · DjangoCon US 2014
• 47mBuilding Scalable Web AppsDavid Cramer · EuroPython 2011
• 21mHigh-Availability DjangoFrankie Dintino · DjangoCon US 2016
• 41mEmbracing the Monolith in Small TeamsLeon Sasson · PyData Chicago 2016
• 56mLarge scale websites performance optimisation tricksGeorgiana Gligor · PHP UK Conference 2016
• 29mSwiss Army Django: Small Footprint ETLNoah Kantrowitz · PyCon AU 2023