[doc] Improve the detailed explanation about how LeoFS works

[mocchira]

Draft - How it works

• [ ] Cluster - leofs_docs/issues/29
  • [ ] Network Toporogy
  • [ ] Monitor
  • [ ] Recover
  • [ ] Update (Attach|Detach)
    • [ ] Rebalance
    • [ ] Takeover
• [ ] Storage
  • [ ] Raw File Format
    • [ ] Object Storage - AVS
    • [ ] Metadata Storage
      • [ ] LevelDB
    • [ ] Mnesia (used by only managers
  • [ ] Large Object
    • [ ] Children
    • [ ] Grand Children
    • [ ] Caveats
      • [ ] leo_gateway settings affect the storage layout
      • [ ] Protocol used by PUT affect the storage layout
      • [ ] Multiple leo_gateways with different mode|chunked_obj_size
  • [ ] ACID
    • [ ] MVCC
    • [ ] When file:fsync called (filed at https://github.com/leo-project/leofs/issues/573
  • [ ] Compaction
    • [ ] Concurrent
    • [ ] Pararrel
    • [ ] little bit STW
  • [ ] Detect corruption
• [ ] Eventual Consistency
  • [ ] Replication
    • [ ] Basic
    • [ ] Rack-aware
  • [ ] Conflict Resolution
  • [ ] How to handle incomplete writes
    • [ ] No rollback: write returns replication error
    • [ ] Hinted Hand Off (>= 1.3.2)
      • [ ] https://quabase.sei.cmu.edu/mediawiki/index.php/Writes_with_Unavailable_Replicas
      • [ ] https://github.com/leo-project/leofs/issues/561
• [ ] Anti Entropy
  • [ ] Active
    • [ ] Read Repair
  • [ ] Passive
    • [ ] Recover (File|Node|Cluster|Ring)
    • [ ] Scrub Tool (filed at https://github.com/leo-project/leofs/issues/574
• [ ] Cache on leo_gateway
  • [ ] Memory
  • [ ] Disk
• [ ] Protocol
  • [ ] External
    • [ ] S3|REST between client and leo_gateway
    • [ ] NFS between client and leo_gateway
  • [ ] Internal
    • [ ] RPC between leo_gateway and leo_storage
    • [ ] RPC between leo_storages
    • [ ] Rebalance between leo_storages
    • [ ] Erlang RPC Limitation
    • [ ] Multi-DC replication between DCs
• [ ] Data Flow
  • [ ] PUT/write
    • [ ] PUT via REST|S3
    • [ ] Write via NFS
  • [ ] GET/read
    • [ ] GET via REST|S3
    • [ ] Read via NFS
    • [ ] Readdir(|plus3) via NFS or listing objects via S3
• [ ] Watchdog
  • [ ] CPU(Load)
  • [ ] Disk I/O(Bandwidth)
  • [ ] Rex
  • [ ] Erlang Internal I/O
  • [ ] Cluster
• [ ] Security
  • [ ] Authentication
  • [ ] Access Control
  • [ ] Audit
    • [ ] Operation Log
• [ ] Multi tenancy
  • [ ] Bucket
  • [ ] User (requested at https://github.com/leo-project/leofs/issues/557

[mocchira]

@yosukehara @windkit Totally refactored and covered as much topics as possible based on my understanding. feel free to add|modify contents if you find something missing|wrong.

[mocchira]

Updated based on FB from Vladimir on ML.

• Add file:fsync on ACID section
• Add scrub tool on Anti Entropy section