1 files changed, 89 insertions, 0 deletions
diff --git a/docs/datastores.md b/docs/datastores.md
index e69de29..ef22c72 100644
--- a/docs/datastores.md
+++ b/docs/datastores.md
@@ -0,0 +1,89 @@
+# Data Stores
+
+## Scope
+
+This standard prescribes database and data storage technologies used to solve many related data-retention concerns.
+The solutions recommended below are designed to encourage deep expertise in a few stable and well-understood systems, rather than maximal "fit" for each distinct use case.
+
+As such, the solutions may not be the most optimal but their performance, maintenance, optimizations, and reliability requirements are understood and supported by the engineering community.
+
+## Terms
+
+- _Database_: provides long-term, durable storage for data whose loss or unavailability would mean violating an application's Availability or Business requirements.
+- _Cache_: provides short-term, volatile storage which does not preserve data.
+  Caches are not in the scope of this standard.
+
+## Capability Matrix
+
+| Capabilities | RDBMS | KV Store | File/Object |
+|--------------|-------|----------|-------------|
+| [Relational]            | X | O |   |
+| [Key-Value]             | X | X | X |
+| [Document-Oriented]     | X | X | X |
+| [Object-Based]          | X | X | X |
+
+_O_: While KV-stores cannot store relational data, some KV-focused databases provide relational-like "tagging" and other attribute aggregations.
+
+## Selection Criteria
+
+### PostgreSQL
+
+Applications SHOULD use PostgreSQL (Aurora in Cloud environments and the latest stable release in OnPrem environments.)
+
+PostgreSQL across all environments supports all storage methods including:
+
+- Simple [Key-Value] stores (via [hstore])
+- [Document-Oriented] storage and queries (via [jsonb])
+- [Large-object] storage directly within the database
+
+If the application is running in the Cloud environment and needs a total data size over [64TB (RDS)][1] or [128TB (Aurora)][2], then it MUST use another approved option.
+
+### DynamoDB
+
+If an Application is hosted in AWS and requires many of:
+
+- Flexible, schemaless data model that will change in the future
+- Read-heavy access model for items
+- Single-millisecond response times
+- Very fast caching for hot keys and values (< 10ms)
+- Multi-region deployments
+
+and does not require any of:
+
+- Strongly consistent reads and writes in all situations
+- Item/Row sizes over 400 KB
+- A maximum data size of over 1TB
+- Joins between different stored values
+
+then it MAY use DynamoDB.
+
+### S3 / File store
+
+If the Applications persists data with many of the following:
+
+- Large files (>100MB each)
+- Large volumes of data (>1 million records)
+- Simple identification requirements (e.g., "filename as key")
+- Simple relational requirements (e.g., folders and files)
+
+and does not require any of the following:
+
+- Low-latency access (< 200ms)
+- Relational join operations
+- Low response times (< 1000ms)
+
+then it MAY use a filestore or S3.
+
+[Aurora (PostgreSQL)]: https://aws.amazon.com/rds/aurora/postgresql-features/
+[RDS (PostgreSQL)]: https://aws.amazon.com/rds/postgresql/
+[DynamoDB]: https://aws.amazon.com/dynamodb/
+[S3]: https://aws.amazon.com/s3/
+[Relational]:https://en.wikipedia.org/wiki/Relational_database
+[Key-Value]: https://en.wikipedia.org/wiki/Key-value_database
+[Document-Oriented]: https://en.wikipedia.org/wiki/Document-oriented_database
+[Object-Based]: https://en.wikipedia.org/wiki/Object_storage#Cloud_storage
+[hstore]: https://www.postgresql.org/docs/13/hstore.html
+[jsonb]: https://www.postgresql.org/docs/current/datatype-json.html
+[Large-object]: https://www.postgresql.org/docs/13/largeobjects.html
+[1]: https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/CHAP_Storage.html
+[2]: https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/CHAP_Limits.html