Docs Request: Compare Average Latency & Cost about open-next HOT 4 OPEN

I work on a medium-sized app. open-next and serverless-nextjs are so cheap that Vercel is hard to justify. Since costs scale in a linear fashion, based on my projections, it would cost around $50 per 10 million requests if using CloudFront and Lambda@Edge. But it really depends on your infrastructure or your website.

As for latency, nothing is stopping you from deploying the same application to Vercel and AWS. I've found that they're pretty similar, at least in my experience, since the functions are hit constantly.

One thing that Vercel definitely has the upper hand on is deployment speed (time from merge or npx sst deploy to live.) This is because, if you're using Lambda@Edge at least, updating CloudFront can take a long time. The problem is less relevant if you're using normal Lambda with function URLs.

Another benefit of Vercel (and seed, sst's deployment and management platform) is pull request previews.

from open-next.

To add to @juliocorzo , you can also setup your own PR preview by setting up a GH Actions and deploying to --stage <SHA> or branch_name

With the new experimental streaming feature, TTFB is in the single digit ms so it's extremely responsive.

from open-next.

@bestickley: Based on my personal experience, the page load time depends on the amount of traffic your website receives. The initial page load time (or "cold start") typically has an overhead of 2 seconds for me. However, if you have consistent traffic, subsequent requests tend to be faster. Additionally, the page load time can be influenced by whether or not you cache the response. When combined with Cloudfront and an efficient caching policy, page responses can be served from a cached version at Cloudfront in less than 250ms. Most companies use Vercel directly, but cdn costs could be very high depending on the use-case.

from open-next.

@martg0, thank you for sharing your experience! That is very helpful.

from open-next.