#CaseStudy - the Man Group moved from managing one of Europe's largest MongoDB clusters to a truly #serverless architecture with #ArcticDB - no more chasing space, no more annoying upgrades - just a software library connecting directly to S3.

How do you build a high-performance database without a server? Alex Seaton breaks down the hard-won lessons from the transition:
• Clock drift is real
• Sometimes you have to redesign from scratch
• CRDTs are useful but very subtle things

🎬 Watch the #InfoQ video and dive into the full architecture behind this shift ⇨ https://bit.ly/3Z9AiwX

📄 #transcript included

#DataEngineering #Database #DatabaseDesign #CRDTs #Python

Meet #ArcticDB 👉 https://bit.ly/4dSp59x

This #InfoQ video explores the practicalities of building a performant time-series datastore and why transactions, particularly the Isolation in ACID (Atomicity, Consistency, Isolation and Durability) is just not worth it.

Curious to learn more? Watch the full video now! #transcript included

#Database #Serverless #Python #DataEngineering

A post from #ArcticDB about using #LMDB on tmpfs for ephemeral DBs where persistence doesn't matter, just performance. https://docs.arcticdb.io/tutorials/lmdb_and_in_memory/

This is stupid. Just use LMDB on a regular filesystem, and run it with MDB_NOSYNC. In that case most I/Os will only get as far as the filesystem cache anyway, so it will be the same perf as tmpfs.

Using tmpfs for LMDB is dumb because RAM allocated there is stuck. While RAM in fscache can still be repurposed if needed.