In contrast, AWS Lambdas, which run on Firecracker, have sub-second startup latency, often just a few hundred milliseconds.

Is there anything comparable on GCP that achieves similar low latency cold starts?

We had a similar bottleneck building out our sandbox routing layer, where we were doing a lookup to a centralized db to route the query. We found that even with a fast KV store, that lookup still added too much overhead. We moved to encoding the routing logic (like region, cluster ID, etc) directly into the subdomain/hostname. This allowed to drop the db read entirely on the hot path and rely on Anycast + latency-based DNS to route the user to the exact right regional gateway instantly. Also, if you ever find yourselves outgrowing standard HTTP proxies for those long-lived agent sessions, I highly recommend looking at Pingora. It gave us way more control over connection lifecycles than NGINX.

For the compute aspect doing sandbox pooling is cool but might kill your unit economics, especially if at some point each tenant has different images. Have you looked into memory snapshots (that way you only have storage costs not full VMs)?

The lesson about "delete code to improve performance" resonates. I've been down similar paths where adding middleware/routing layers seemed like good abstractions, but they ended up being the performance bottleneck.

A few thoughts on this approach:

1. Warm pools are brilliant but expensive - how are you handling the economics? With multi-region pools, you're essentially paying for idle capacity across multiple data centers. I'm curious how you balance pool size vs. cold start probability.

2. Fly's replay mechanism is clever, but that initial bounce still adds latency. Have you considered using GeoDNS to route users to the correct regional endpoint from the start? Though I imagine the caching makes this a non-issue after the first request.

3. For the JWT approach - are you rotating these tokens per-session? Just thinking about the security implications if someone intercepts the token.

The 79ms → 14ms improvement is night and day for developer experience. Latency under 20ms feels instant to humans, so you've hit that sweet spot.

This article reads like a thinly veiled ad. Certainly not the best way to start a technical blog. If you didn't have the technical insight to know that physics is a factor in latency, why should I trust you with the problems your product actually solves?

I also had switched a head of line service call that was, for reasons I never sorted out, costing us 30ms TTFB per request for basically fifty bytes of data, to use a long poll in Consul because the data was only meant to be changed at most once every half hour and in practice twice a week. So that latency was hidden in dev sandbox except for startup time, where we had several consul keys being fetched in parallel and applied in order, so one more was hardly noticeable.

The nasty one though was that Artifactory didn’t compress its REST responses, and when you have a CI/CD pipeline that’s been running for six years with half a hundred devs that response is huge because npm is teh dumb. So our poor UI lead kept having npm install timeout and the UI team’s answer for “my environment isn’t working” started with clearing your downloaded deps and starting over.

They finally fixed it after we (and presumably half of the rest of their customers) complained but I was on the back 9 of migrating our entire deployment pipeline to docker and so I had nginx config fairly fresh in my brain and I set them up a forward proxy to do compression termination. It still blew up once a week but that was better than him spending half his day praying to the gods of chaos.

Valuable insight /s