25 million+ Low Latency IP lookups/month on Infinitely Scalable Serverless

25 million+ Low Latency IP lookups/month on Infinitely Scalable Serverless(highscalability.com)

6 points by mitgraduate 7 years ago

zzzcpan 7 years ago |

https://news.ycombinator.com/item?id=16745376 (29 comments)

crb002 7 years ago |

Or just use SQLite backed by S3. Most latency is to the data center. No need for DynamoDB billing overhead. Cache 5mb or so of most frequent static data in the Lambda.

mitgraduate 7 years ago | |

Can you please explain how we can access sqlite database hosted by s3 in lambda?

I googled it but it seems s3 is only for hosting static files and it will not be possible to directly query Sqlite DB on s3.

crb002 7 years ago | | |

SQLite is just a flat file database format. You can use stock Rust, Python, C++ SQLite client. Probably want to shard the SQLite files to prevent hot spots on write locks. Use a quadtree for geospatial sharding.

crb002 7 years ago | |

Writes are the only reason you need DynamoDB. Even then you can replicate to S3 so reads remain cheap.

Only issue is transactional writes, but those can wait for S3 write confirmation in the lock.

For example bank statement, ok to read from S3 SQLite file. Actual credit/debit you must use locking.