The more I use serverless, the more I want to use servers, it feels like I’m just building fixes to make serverless work
But it’s hard to achieve such low latencies globally like
@cloudflare
workers
12
3
82
Replies
Might be related
1
1
9
@chronark_
@Cloudflare
After a threshold, latency becomes a vanity benchmark.
That threshold is easy to achieve with a multi-region deployment setup. (fly is a good example here)
2
0
6
@ImSh4yy
@Cloudflare
I had really bad latency on fly cause they’d frequently route requests all over the world even when there are idle replicas close by
But as we are sitting in the critical path of our customers, latency is far from a vanity benchmark, it’s critical for us to add as little…
4
0
2
@chronark_
@Cloudflare
Yes i get the same feeling. Life would be so easy if I had a persistent server. On the other hand, i’d need to take care of scaling up and down and then I just don’t have time for it.
My take away: serverless great to get started, server + devops great for scaling
1
0
2
@hichaelmart
@Cloudflare
lack of persistent state, broken js tooling and half baked products from cloudflare
and yes I know of durable objects and caching, we use both, it's still a pain
1
0
0
@wunderkuh
@Cloudflare
we had horrible latencies on fly due to bad routing around the world even though there were idle machines close by
0
0
0
@chronark_
@Cloudflare
Just buy bare metal nodes in all the global edge locations and run your applications there behind geodns
1
0
0
@chronark_
@Cloudflare
It’s hard to achieve something similar. Cloudflare is great at optimising its systems.
You could check out having a global deployment with AWS Global Accelerator and EC2 across the 33 AWS regions.
1
0
0
@chronark_
@Cloudflare
Also the constant fighting with the lack of proper runtime when compatibility mode breaks. A year ago I was huge proponent of CF workers, now I've gradually switched back to docker.
0
0
1