Railway vs Heroku
My team have been long time heroku customers on a decent sized app (we spend about $3k/mo for service) and we're looking for alternatives. Ever since Salesforce bought heroku their service has... well... degraded into Salesforce quality π . I'm loving railway so far but I'm looking for some more technical details about how railway compares to heroku. For example:
1. Does railway rely on AWS?
2. Is it edge based (like say vercel)?
3. Is there an official performance comparison between railway and heroku somewhere?
4. Is the postgres integration vanilla postgres or is it customized like heroku with a lot of sugar?
5. Are there large apps running stably in production on railway?
9 Replies
1. We rely on GCP
2. We're not edge based. We run on full fledged datacenters, which means there's no restrictions on workloads (runtime limits, payload size, etc)
3. No official numbers but we've found users who switch from places like Heroku and Render usually see increased response times (sometimes up to 3x but nothing official). But that's anecdotal; Mostly, all these platforms will give you similar performance. Railway's differentiator is the experience, which lets your engineers ship orders of magnitude faster.
4. It's vanilla but it's your own. Feel free to add whatever you want (you can psql in etc). Anything that's a hard requirement?
5. We've got over 75k+ people on the platform ATM, ranging from early stage startups to larger companies, using it for a variety of applications. You can read more under "Usecases" (https://docs.railway.app/reference/usecases).
1. π
2. π good we prefer datacenters
3. I wonder if this is just because the person was closer to which datacenter railway happened to put them on, as opposed to it actually being faster.
4. Roping in my colleague who can speak more to our needs re: postgres
5. π
3. Measured via APM response time so it actually only counts from when the request hits the server
Unknown Userβ’2y ago
Message Not Public
Sign In & Join Server To View
1. We don't have a native primitive but people use things like https://www.snaplet.dev/ in the past and I'm told it's quite good
2. To upgrade major versions with zero downtime, you'll want to setup another instance as a read replica, then do a failover. We currently don't have an automated process for this
3. This can be achieved by adding another service, which runs as a cron, and runs the commands required to do a DB backup (and then emit it to somewhere like a backblaze bucket). I've personally been meaning to write a template for this, but haven't had the time.
Snaplet
Snaplet gives developers production-accurate data and ephemeral databases they can code against, so they can focus on shipping. Ditch the seed script forever!
The way Railway is setup is that it's more of a "mesh of things".
So instead of a platform offering one standard way to do backups, which might be in the wrong format, or storing things at the wrong cadence, you can simply deploy a service to set this up however you'd like. Our goal is to make the setting up of those services as fast as possible.
Unknown Userβ’2y ago
Message Not Public
Sign In & Join Server To View
But of course! We're here to help with anything.
Additionally, when you sign up for a teams plan, it'll create a direct, private line between our team and yours
So, if you have any questions, want to do an onboarding call, we're happy to do that over Discord (or Zoom/Meet/whatever)
No pressure; we've designed this to be completely self serve and as async as possible π
And no need to sign up for a teams plan right away; use the free/dev account to get started and once you're happy, create a team and you can just simply drag the projects from your account, to the team account
Just an FYI I wrote a lil guide here as well mentioning the key benefits Railway has over Heroku
Might be helpful for y'all
https://www.notion.so/railwayapp/Railway-vs-Heroku-ed0b849218ed47b6ad01e6ded619b648