Storage Node / Cluster - HALO-DB spec

A Storage Node (or cluster) stores and distributes IPLD blocks.

many backends are possible (e.g. filesystem, object store, tape?)
blocks are accessible via bitswap and HTTP trustless gateway
advertises available CIDs using bitswap, DHT and IPNI announcements
exposes a pinning service (users can ask the node to keep certain blocks)
- via pinning API
- via HTTP upload of CAR archives
- authenticated by Auth Service
exposes storage attestation API, a HTTP API answering questions like
- do you store this CID? -- and it’s children?
- how long will you keep it?
(TBD / maybe) follows a remote pinlist (e.g. as in ORCESTRA pinlist.yaml

Additional Services¶

These are likely very useful, but it’s likely ok to have fewer replicas of these.

Network Indexer (tracks, indexes & publishes IPNI updates)
Delegated Routing Service (exposes routing information from DHT & Network Indexer via HTTP)
Standalone HTTP Gateway (maybe less relevant in the longer term, exposes IPFS content via HTTP)

specify requirements for the above mentioned services more thoroughly
develop quota management (or something that ensures all participants can use the service without becoming unfair)
specify security recommendations to operate the services in a safe manner
implement storage node as blueprint for the infrastructure available to project partners
install & operate such nodes at multiple sites
install & operate (at least) one network indexer
install & operate (at least) one delegated routing service
(TBD / maybe later) develop system for transparent long-term archiving (e.g. Tape connector)