undefined | Better HN

0 pointsmrkurt5y ago0 comments

Oh hello! (disclosure, I work on Fly.io).

We're more like Fargate than Lambda, but Fly apps scale and balance across regions. I wouldn't really call it "more edgy", it's just built specifically for running app servers close to users (vs emulating a traditional single region datacenter).

We actually started with a JS runtime like CloudFlare has, but it was wrong for our customers. They're better off just running Docker images.

There's a lot more in our HN launch thread: https://news.ycombinator.com/item?id=22616857

0 comments

3 comments · 1 top-level

rbflx5y ago· 2 in thread

Sorry if this is somewhat unrelated to the main thread, but considering that someone working on Fly.io is here, I wanted to ask some questions that I couldn't find the answers to in your documentation.

1. What's the average cold start time when spinning up a new instance? For comparison, AWS Fargate often has cold starts of around 1 minute.

2. Is there a limit on the container size created by the Dockerfile?

3. Is there a limit on the number of concurrent instances a single user can run?

mrkurtOP5y ago

1. Cold start times are wildly variable. They can be <1s for optimized, fast booting docker images. We actually mask cold start time as much as possible, though, so you'll almost never see one.

2. There's no hard limit. We've had 8GB images running (that I know of). Cold starts can be pretty bad on those just because pulling down an 8gb filesystem is slow.

3. Also no hard limit. I'm curious what you're doing though! Feel free to email me: kurt at fly.io

sudhirj5y ago

I'm using Go, so the default no-config container is about 140MB. But it's possible to use a multi step build to bring that down to about 10MB (build it in a full container and copy into a very lightweight scratch container). That's in the range of Lambda or CloudFlare worker sizes, so startup speeds shouldn't be an issue.

j / k navigate · click thread line to collapse