What is the plan for PDP data scaling with the sun...
# permit-saas
s
What is the plan for PDP data scaling with the sunsetting of PDP sharding tomorrow?
a
Hi @Sam Jakos, I'm an AI assistant for Permit.io. I'll analyze available sources and get back to you shortly.
* The PDP Sharding feature is indeed being deprecated and will be sunsetted on September 1st, 2025 (1) . The documentation states that Permit.io is "taking other and a more flexible and scalable approach for dealing with high volume" , though specific details about the replacement approach aren't provided in the available sources. For current scaling options while waiting for the new approach, you have several alternatives: ​*Available Scaling Patterns:*​ 1. ​*Sidecar Deployment*​ - Deploy PDPs as sidecars next to each of your services for near-zero latency (2) 2. ​*Centralized PDP*​ - Use a single PDP to handle authorization queries from multiple services (simpler but can become a bottleneck) 3. ​*Cluster of PDPs*​ - Deploy multiple PDPs behind a load balancer for high availability and scalability . You can use the PDP Helm chart for Kubernetes deployments 4. ​*PDP Caching*​ - The PDP includes a configurable caching mechanism that can use either in-memory storage for single instances or Redis for distributed caching across multiple PDPs The documentation indicates that sharded PDPs are currently only available for early-access-program customers (1) , so if you're currently using sharding, I'd recommend reaching out to help@permit.io to discuss migration plans and learn about the new scaling approach that will replace sharding.
To learn more, see the sources I considered: [[1] Use Data in Scale | Permit.io Documentation](https://docs.permit.io/how-to/manage-data/pdp-sharding) [[2] Permit's PDP (Policy Decision Point)](https://docs.permit.io/concepts/pdp/overview)
s
Is it possible to deploy PDPs that only sync a specific Tenant’s facts? @permit support automation2
o
Existing customers using the feature are getting long term support. We are very close to release a new version of the PDP that will include new storage options.
👍 1
You probably meant to tag @Ask Inkeep
a
Thanks for tagging me @Or Weis. I'll look through relevant sources and get back to you shortly.
Currently, the available documentation doesn't describe a built-in feature for deploying PDPs that sync only a specific tenant's facts. (1) (2) (3) The PDP Sharding feature (which is being deprecated) was designed to split data across multiple PDPs with each shard responsible for a subset of tenants using consistent hashing. (4) However, this approach divides tenants automatically across shards rather than allowing you to explicitly configure a PDP for a single specific tenant. In standard deployments, PDPs sync with a specific environment via the environment API key, and environments contain all tenants within them. (2) (3) Tenants are part of the facts (data) layer that gets synchronized to the PDP. (1) For custom requirements like tenant-specific PDP deployments, I'd recommend reaching out to help@permit.io to discuss your specific use case, especially given the upcoming changes to data scaling approaches mentioned in the thread.
To learn more, see the sources I considered: [[1] Multi-Tenant Authorization](https://docs.permit.io/concepts/multi-tenant-authorization) [[2] Concepts / Glossary](https://docs.permit.io/overview/glossary) [[3] Permit's PDP (Policy Decision Point)](https://docs.permit.io/concepts/pdp/overview) [[4] Use Data in Scale | Permit.io Documentation](https://docs.permit.io/how-to/manage-data/pdp-sharding)
o
What is the scale you're looking at the moment?
s
The scale is relatively small, but I’m looking into how I could support Tenant data segregation potentially in a hybrid hosting model.
o
This is planned to be addressed in the new PDP storage release, for now my beat recommendation would be to use different Permit environments
s
Do you have an ETA on the new PDP?
o
This feature is the bane of my existence at this point 😅 - but I'm confident by end of October if not sooner it will be released
😂 1
CC: @Omer Zuarets, @Dan
o
Hey @Sam Jakos, The deprecated PDP Sharding doesn’t actually give you a “promise” that each PDP holds only 1 tenant - even having 1<>1 number of tenants to number of PDP shards it isn’t guaranteed by the hash ring, and the overhead of having 1<>1 PDP shards to tenants is really big besides of it being deprecated. Currently, all the tenants’ data are stored on the same PDP - even with the new storage layers for the PDP it would be pretty similar. The data is separated logically in the PDP so each tenant data is decoupled for the others meaning there are no connections between tenant objects. You can think of it similarly to a single RDS instance holding all your tenants data and the separation between them is done logically in the application via databases/schemas/table columns
o
Omer I think Sam cares more about the upcoming solution at this point 😇