Reposted by Tim Caswell
We’ve made all CDN serving on Vercel 10% faster (P75), while reducing memory by 15%.
Learn the behind the scenes of applying Bloom filters to this problem, while serving 1T monthly requests.
vercel.com/blog/how-we...
vercel.com
How we made global routing faster with Bloom filters - Vercel
We replaced slow JSON path lookups with Bloom filters in our global routing service, cutting memory usage by 15% and reducing 99th percentile lookup times from hundreds of milliseconds to under 1 ms. Here’s how we did it.
![export function eytzinger<T>(arr: T[]): T[] {
const n = arr.length
const tree = new Array(n)
let i = 0
function build(k: number): void {
if (k > n) return
build(2 * k)
tree[k - 1] = arr[i++]
build(2 * k + 1)
}
build(1)
return tree
}](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:evdiwz3kr7rtta3j67vzxs27/bafkreifrhhtum37wcfhbjwz7njszwvu56s2zsmfacrz6xx3ogbqspzwktm)
![let log = [
[0, ['src', 'answer.js'], [[0], [0]], 'export let answer = 42\n'],
[1, ['src', 'index.js'], [[1], [0, 1]], 'import { answer } from ...'],
[2, ['README.md'], [[2]], 'This *is* the README\n'],
[3, ['src', 'more.js'], [[2, 3], [0, 1, 3]], 'export let more = ...'],
]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:evdiwz3kr7rtta3j67vzxs27/bafkreidj4a2tgva4jotv3ywevjmcctcvzdprzgjj4g74subtatxtjfqvme)
