Heroku Postgres plans offer resource tuning (adjustable CPU, Memory, disk, etc.)

[johlym]

Required Terms

• [X] I agree to follow this project's Code of Conduct
• [X] I have read and accept the Salesforce Program Agreement

What service(s) is this request for?

Heroku Postgres

Tell us about what you're trying to solve. What challenges are you facing?

As the datasets our customers manage on Heroku Postgres grow, one concern is the disproportionate sizing when scaling to larger plans. In some cases, said datasets may grow substantially, but the computational needs may not, and thus paying for orders of magnitude more resources than necessary (ex: copious amounts of unused RAM and compute when only extra disk will do just fine for a particular workload).

This can create a situation where users feel they're paying for resources they don't need or won't use, and start to see the value in the platform diminish as they find alternatives elsewhere that offer more fine-grained resource allocation.

[jessehanley]

@johlym thanks for opening this issue.

The inflexibility of Heroku's Postgres plans is one of the key reasons we, as an Enterprise customer, have begun looking at alternatives. Would love to see this expedited and shipped within the coming months before our renewal 😄

[afawcett]

Thanks for raising @johlym and commenting @jessehanley.

Good news - we agree you - we have raised another issue that contains further perspective from us on this - https://github.com/heroku/roadmap/issues/60. We are currently scoping this out to better understand how bring this to our customers asap.

[jessehanley]

Hey @afawcett,

Returning to this as there hasn't been much action on this? Reading @jbrown-heroku post it seems Heroku isn't listening to the ask from customers. @johlym's issue here, which I agree with, is that customers are paying for more than they need in the compute dept.

I'm personally looking at going from a Standard-6 to Standard-7 because I need more storage and looking at a whopping $1.5k increase for the extra 500GB? We need elasticity at all levels.

[jbrown-heroku]

We are working on ways to allow for more flexibility around our data plans. We understand the mismatch with disk vs compute/memory per customer's needs.

[jessehanley]

Any progress here?

[aldericon]

We have the issue of needing more RAM but not more storage capacity. Is this something that will be resolved here? Do we have an update on when this will go out?

[jessehanley]

@aldericon in ticket #60 (closed) they've decided they won't address it

(╯°益°)╯彡┻━┻

[jbrown-heroku]

We are going to address the "mismatch" of resources in our current offerings. I cannot share ETA yet. I understand this is an important change to help many customers to right-size their database.

[capeterson]

@aldericon under the hood our plans are still coupled to AWS instance sizes, which gives us less flexibility on independent scaling of memory and CPU than we'd like, so our current roadmap won't cover being able to scale memory as independently as you (or I) would like.

We put a heavy focus on storage scaling since it's an issue we hear constantly across virtually all forms of customer feedback. I'd very much like to give you more flexibility on a more flexible per-resource basis if we're able to, but I also want to be transparent that it's dependent on finding some creative ways to solve underlying infra size challenges.

With all that said, I'd love to hear more about you use case so we can build a better narrative around the value of independent memory scaling.

@jessehanley I do want to explicitly confirm your example of scaling up your DB just for more storage storage is very much something we are going to solve by making additional storage available on a per-GB, per-month basis when we launch our Performance tier on AWS Aurora. We're targeting (normal disclaimers apply about any forward looking statements, especially dates) having that solution available sometime between Q4 this year and Q2 next year in our current plan.

More feedback is very welcome as always. I hope the extra detail is at least useful for the sake of your planning needs.

[aldericon]

@capeterson Thanks for the reply. When I did performance testing late last year, I found that we'd benefit from having more RAM, since our website is mostly dynamic (forums). However I've spent a lot of time the last few weeks finding slow points in the system and improving on it. If we'd able to pay for more storage space in the future without quite the jump that we currently have (going from Standard-0 to Standard-2 is a lot for us), then I'll continue finding other ways to speed the site up.

[jessehanley]

@capeterson I read about this on Twitter, slightly concerned re: Aurora's pricing model vs. say a Heroku RDS.

If Heroku launches Heroku Postgres Performance will the migration be seamless from the current Heroku offering?

[jbrown-heroku]

Thank you @jessehanley for your reply. We understand the fear around possible “run away cost” with Aurora model. This is why we are focused on delivering flexibility around storage without compromising predictable pricing for the compute or IO. Your feedback is valuable and we will address the concerns the best we can. And yes, we want to make the migration as seamless as possible and we are investing heavily into improving the experience all around, from migration to upgrades to storage growth. These are in our current plans.

[jbrown-heroku]

Sharing the pilot program waitlist form: https://forms.gle/qLH9Nui4xxAzhYCY7 Please signup if you would like to test the pilot version of our newest database. This is a great opportunity to work closely with Heroku Product and Engineering teams on a brand new product, and shape it for the future. We will go through a selection process and reach around January 2025 with instructions if you are selected.