You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardexpand all lines: README.md
+18
Original file line number
Diff line number
Diff line change
@@ -8,3 +8,21 @@ The Honu Database is an eventually consistent replicated document database that
8
8
Smart anti-entropy uses reinforcement learning with multi-armed bandits to optimize replication. Adaptive consistency reduces costs (ingress and egress data transfer) as well as improves consistency by lowering the likelihood of stale reads or forked writes.
9
9
10
10
The goal of the database is to provide scalable data retrieval both in terms of number of nodes (e.g. scale to 100s of nodes) and amount of data (hundreds of terabytes). In addition to scale, this database provides data access controls, privacy and provenance, and other security related features. In short, HonuDB is a distributed data governance database for machine learning and artificial intelligence workloads.
11
+
12
+
13
+
## Object Storage
14
+
15
+
Protocol Buffers are a compact, cross-language compatible data serialization format that facilitates compact network communications. However, in order to make them general purpose and flexible, they require a lot of reflection to work in Go. Since a database is a high performance application, I've implemented a data serialization format that uses no reflection and as a result performs far better at decoding than Protocol Buffers:
0 commit comments